The companies’ efforts follow the announcement of OpenAI’s latest LLMs, including the o1 generative pre-trained transformer model backed by Microsoft. The model is intended to tackle difficult tasks, paving the way for advances in science, coding, and mathematics.

During the conference, Kunal Zhilin, founder of Moonshot AI, underlined the importance of the o1 model, adding that it has the potential to reshape various industries and create new opportunities for AI startups.

Zhilin stated that reinforcement learning and scalability might be pivotal for AI development. He spoke of the scaling law, which states that larger models with more training data perform better.

“This approach pushes the ceiling of AI capabilities,” Zhilin said, adding that OpenAI o1 has the potential to disrupt sectors and generate new opportunities for startups.

OpenAI has also stressed the model’s ability to solve complex problems, which it says operate in a manner similar to human thinking. By refining its strategies and learning from mistakes, the model improves its problem-solving capabilities.

Zhilin said companies with enough computing power will be able to innovate not only in algorithms, but also in foundational AI models. He sees this as pivotal, as AI engineers rely increasingly on reinforcement learning to generate new data after exhausting available organic data sources.

StepFun CEO Jiang Daxin concurred with Zhilin but stated that computational power remains a big challenge for many start-ups, particularly due to US trade restrictions that hinder Chinese enterprises’ access to advanced semiconductors.

“The computational requirements are still substantial,” Daxin stated.

An insider at Baichuan AI has said that only a small group of Chinese AI start-ups — including Moonshot AI, Baichuan AI, Zhipu AI, and MiniMax — are in a position to make large-scale investments in reinforcement learning. These companies — collectively referred to as the “AI tigers” — are involved heavily in LLM development, pushing the next generation of AI.

More from the Apsara Conference

Also at the conference, Alibaba Cloud made several announcements, including the release of its Qwen 2.5 model family, which features advances in coding and mathematics. The models range from 0.5 billion to 72 billion parameters and support approximately 29 languages, including Chinese, English, French, and Spanish.

Specialised models such as Qwen2.5-Coder and Qwen2.5-Math have already gained some traction, with over 40 million downloads on platforms Hugging Face and ModelScope.

Alibaba Cloud added to its product portfolio, delivering a text-to-video model in its picture generator, Tongyi Wanxiang. The model can create videos in realistic and animated styles, with possible uses in advertising and filmmaking.

Alibaba Cloud unveiled Qwen 2-VL, the latest version of its vision language model. It handles videos longer than 20 minutes, supports video-based question-answering, and is optimised for mobile devices and robotics.

For more information on the conference, click here.

(Photo by: @Guy_AI_Wise via X)

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post OpenAI’s latest LLM opens doors for China’s AI startups appeared first on AI News.

The expectations of results from AI are balanced at present with real-world realities. And there remains a good deal of suspicion about the technology, again in balance with those who are embracing it even in its current nascent stages. The closed-loop nature of the well-known LLMs is being challenged by instances like Llama, DeepSeek, and Baidu’s recently-released Ernie X1.

In contrast, open source development provides transparency and the ability to contribute back, which is more in tune with the desire for “responsible AI”: a phrase that encompasses the environmental impact of large models, how AIs are used, what comprises their learning corpora, and issues around data sovereignty, language, and politics. 

As the company that’s demonstrated the viability of an economically-sustainable open source development model for its business, Red Hat wants to extend its open, collaborative, and community-driven approach to AI. We spoke recently to Julio Guijarro, the CTO for EMEA at Red Hat, about the organisation’s efforts to unlock the undoubted power of generative AI models in ways that bring value to the enterprise, in a manner that’s responsible, sustainable, and as transparent as possible. 

Julio underlined how much education is still needed in order for us to more fully understand AI, stating, “Given the significant unknowns about AI’s inner workings, which are rooted in complex science and mathematics, it remains a ‘black box’ for many. This lack of transparency is compounded where it has been developed in largely inaccessible, closed environments.”

There are also issues with language (European and Middle-Eastern languages are very much under-served), data sovereignty, and fundamentally, trust. “Data is an organisation’s most valuable asset, and businesses need to make sure they are aware of the risks of exposing sensitive data to public platforms with varying privacy policies.” 

The Red Hat response 

Red Hat’s response to global demand for AI has been to pursue what it feels will bring most benefit to end-users, and remove many of the doubts and caveats that are quickly becoming apparent when the de facto AI services are deployed. 

One answer, Julio said, is small language models, running locally or in hybrid clouds, on non-specialist hardware, and accessing local business information. SLMs are compact, efficient alternatives to LLMs, designed to deliver strong performance for specific tasks while requiring significantly fewer computational resources. There are smaller cloud providers that can be utilised to offload some compute, but the key is having the flexibility and freedom to choose to keep business-critical information in-house, close to the model, if desired. That’s important, because information in an organisation changes rapidly. “One challenge with large language models is they can get obsolete quickly because the data generation is not happening in the big clouds. The data is happening next to you and your business processes,” he said. 

There’s also the cost. “Your customer service querying an LLM can present a significant hidden cost – before AI, you knew that when you made a data query, it had a limited and predictable scope. Therefore, you could calculate how much that transaction could cost you. In the case of LLMs, they work on an iterative model. So the more you use it, the better its answer can get, and the more you like it, the more questions you may ask. And every interaction is costing you money. So the same query that before was a single transaction can now become a hundred, depending on who and how is using the model. When you are running a model on-premise, you can have greater control, because the scope is limited by the cost of your own infrastructure, not by the cost of each query.”

Organisations needn’t brace themselves for a procurement round that involves writing a huge cheque for GPUs, however. Part of Red Hat’s current work is optimising models (in the open, of course) to run on more standard hardware. It’s possible because the specialist models that many businesses will use don’t need the huge, general-purpose data corpus that has to be processed at high cost with every query. 

“A lot of the work that is happening right now is people looking into large models and removing everything that is not needed for a particular use case. If we want to make AI ubiquitous, it has to be through smaller language models. We are also focused on supporting and improving vLLM (the inference engine project) to make sure people can interact with all these models in an efficient and standardised way wherever they want: locally, at the edge or in the cloud,” Julio said. 

Keeping it small 

Using and referencing local data pertinent to the user means that the outcomes can be crafted according to need. Julio cited projects in the Arab- and Portuguese-speaking worlds that wouldn’t be viable using the English-centric household name LLMs. 

There are a couple of other issues, too, that early adopter organisations have found in practical, day-to-day use LLMs. The first is latency – which can be problematic in time-sensitive or customer-facing contexts. Having the focused resources and relevantly-tailored results just a network hop or two away makes sense. 

Secondly, there is the trust issue: an integral part of responsible AI. Red Hat advocates for open platforms, tools, and models so we can move towards greater transparency, understanding, and the ability for as many people as possible to contribute. “It is going to be critical for everybody,” Julio said. “We are building capabilities to democratise AI, and that’s not only publishing a model, it’s giving users the tools to be able to replicate them, tune them, and serve them.” 

Red Hat recently acquired Neural Magic to help enterprises more easily scale AI, to improve performance of inference, and to provide even greater choice and accessibility of how enterprises build and deploy AI workloads with the vLLM project for open model serving. Red Hat, together with IBM Research, also released InstructLab to open the door to would-be AI builders who aren’t data scientists but who have the right business knowledge. 

There’s a great deal of speculation around if, or when, the AI bubble might burst, but such conversations tend to gravitate to the economic reality that the big LLM providers will soon have to face. Red Hat believes that AI has a future in a use case-specific and inherently open source form, a technology that will make business sense and that will be available to all. To quote Julio’s boss, Matt Hicks (CEO of Red Hat), “The future of AI is open.” 

Supporting Assets: 

Tech Journey: Adopt and scale AI

The post Red Hat on open, small language models for responsible, practical AI appeared first on AI News.

According to the report, titled ‘Rebooting Copyright: How the UK Can Be a Global Leader in the Arts and AI,’ the global race for cultural and technological leadership is still up for grabs, and the UK has a golden opportunity to take the lead.

The report emphasises that countries that “embrace change and harness the power of artificial intelligence in creative ways will set the technical, aesthetic, and regulatory standards for others to follow.”

Highlighting that we are in the midst of another revolution in media and communication, the report notes that AI is disrupting how textual, visual, and auditive content is created, distributed, and experienced, much like the printing press, gramophone, and camera did before it.

“AI will usher in a new era of interactive and bespoke works, as well as a counter-revolution that celebrates everything that AI can never be,” the report states.

However, far from signalling the end of human creativity, the TBI suggests AI will open up “new ways of being original.”

The AI revolution’s impact isn’t limited to the creative industries; it’s being felt across all areas of society. Scientists are using AI to accelerate discoveries, healthcare providers are employing it to analyse X-ray images, and emergency services utilise it to locate houses damaged by earthquakes.

The report stresses that these cross-industry advancements are just the beginning, with future AI systems set to become increasingly capable, fuelled by advancements in computing power, data, model architectures, and access to talent.

The UK government has expressed its ambition to be a global leader in AI through its AI Opportunities Action Plan, announced by Prime Minister Keir Starmer on 13 January 2025. For its part, the TBI welcomes the UK government’s ambition, stating that “if properly designed and deployed, AI can make human lives healthier, safer, and more prosperous.”

However, the rapid spread of AI across sectors raises urgent policy questions, particularly concerning the data used for AI training. The application of UK copyright law to the training of AI models is currently contested, with the debate often framed as a “zero-sum game” between AI developers and rights holders. The TBI argues that this framing “misrepresents the nature of the challenge and the opportunity before us.”

The report emphasises that “bold policy solutions are needed to provide all parties with legal clarity and unlock investments that spur innovation, job creation, and economic growth.”

According to the TBI, AI presents opportunities for creators—noting its use in various fields from podcasts to filmmaking. The report draws parallels with past technological innovations – such as the printing press and the internet – which were initially met with resistance, but ultimately led to societal adaptation and human ingenuity prevailing.

The TBI proposes that the solution lies not in clinging to outdated copyright laws but in allowing them to “co-evolve with technological change” to remain effective in the age of AI.

The UK government has proposed a text and data mining exception with an opt-out option for rights holders. While the TBI views this as a good starting point for balancing stakeholder interests, it acknowledges the “significant implementation and enforcement challenges” that come with it, spanning legal, technical, and geopolitical dimensions.

In the report, the Tony Blair Institute for Global Change “assesses the merits of the UK government’s proposal and outlines a holistic policy framework to make it work in practice.”

The report includes recommendations and examines novel forms of art that will emerge from AI. It also delves into the disagreement between rights holders and developers on copyright, the wider implications of copyright policy, and the serious hurdles the UK’s text and data mining proposal faces.

Furthermore, the Tony Blair Institute explores the challenges of governing an opt-out policy, implementation problems with opt-outs, making opt-outs useful and accessible, and tackling the diffusion problem. AI summaries and the problems they present regarding identity are also addressed, along with defensive tools as a partial solution and solving licensing problems.

The report also seeks to clarify the standards on human creativity, address digital watermarking, and discuss the uncertainty around the impact of generative AI on the industry. It proposes establishing a Centre for AI and the Creative Industries and discusses the risk of judicial review, the benefits of a remuneration scheme, and the advantages of a targeted levy on ISPs to raise funding for the Centre.

However, the report has faced strong criticism. Ed Newton-Rex, CEO of Fairly Trained, raised several concerns on Bluesky. These concerns include:

• The report repeats the “misleading claim” that existing UK copyright law is uncertain, which Newton-Rex asserts is not the case.
• The suggestion that an opt-out scheme would give rights holders more control over how their works are used is misleading. Newton-Rex argues that licensing is currently required by law, so moving to an opt-out system would actually decrease control, as some rights holders will inevitably miss the opt-out.
• The report likens machine learning (ML) training to human learning, a comparison that Newton-Rex finds shocking, given the vastly different scalability of the two.
• The report’s claim that AI developers won’t make long-term profits from training on people’s work is questioned, with Newton-Rex pointing to the significant funding raised by companies like OpenAI.
• Newton-Rex suggests the report uses strawman arguments, such as stating that generative AI may not replace all human paid activities.
• A key criticism is that the report omits data showing how generative AI replaces demand for human creative labour.
• Newton-Rex also criticises the report’s proposed solutions, specifically the suggestion to set up an academic centre, which he notes “no one has asked for.”
• Furthermore, he highlights the proposal to tax every household in the UK to fund this academic centre, arguing that this would place the financial burden on consumers rather than the AI companies themselves, and the revenue wouldn’t even go to creators.

Adding to these criticisms, British novelist and author Jonathan Coe noted that “the five co-authors of this report on copyright, AI, and the arts are all from the science and technology sectors. Not one artist or creator among them.”

While the report from Tony Blair Institute for Global Change supports the government’s ambition to be an AI leader, it also raises critical policy questions—particularly around copyright law and AI training data.

(Photo by Jez Timms)

See also: Amazon Nova Act: A step towards smarter, web-native AI agents

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post Tony Blair Institute AI copyright report sparks backlash appeared first on AI News.

But virtual machines (VMs) have several downsides. Often, an entire virtualised operating system is overkill for many applications, and although very much more malleable, scalable, and agile than a fleet of bare-metal servers, VMs still require significantly more memory and processing power, and are less agile than the next evolution of this type of technology – containers. In addition to being more easily scaled (up or down, according to demand), containerised applications consist of only the necessary parts of an application and its supporting dependencies. Therefore apps based on micro-services tend to be lighter and more easily configurable.

Virtual machines exhibit the same security issues that affect their bare-metal counterparts, and to some extent, container security issues reflect those of their component parts: a mySQL bug in a specific version of the upstream application will affect containerised versions too. With regards to VMs, bare metal installs, and containers, cybersecurity concerns and activities are very similar. But container deployments and their tooling bring specific security challenges to those charged with running apps and services, whether manually piecing together applications with choice containers, or running in production with orchestration at scale.

Container-specific security risks

• Misconfiguration: Complex applications are made up of multiple containers, and misconfiguration – often only a single line in a .yaml file, can grant unnecessary privileges and increase the attack surface. For example, although it’s not trivial for an attacker to gain root access to the host machine from a container, it’s still a too-common practice to run Docker as root, with no user namespace remapping, for example.

• Vulnerable container images: In 2022, Sysdig found over 1,600 images identified as malicious in Docker Hub, in addition to many containers stored in the repo with hard-coded cloud credentials, ssh keys, and NPM tokens. The process of pulling images from public registries is opaque, and the convenience of container deployment (plus pressure on developers to produce results, fast) can mean that apps can easily be constructed with inherently insecure, or even malicious components.

• Orchestration layers: For larger projects, orchestration tools such as Kubernetes can increase the attack surface, usually due to misconfiguration and high levels of complexity. A 2022 survey from D2iQ found that only 42% of applications running on Kubernetes made it into production – down in part to the difficulty of administering large clusters and a steep learning curve.

According to Ari Weil at Akamai, “Kubernetes is mature, but most companies and developers don’t realise how complex […] it can be until they’re actually at scale.”

Container security with machine learning

The specific challenges of container security can be addressed using machine learning algorithms trained on observing the components of an application when it’s ‘running clean.’ By creating a baseline of normal behaviour, machine learning can identify anomalies that could indicate potential threats from unusual traffic, unauthorised changes to configuration, odd user access patterns, and unexpected system calls.

ML-based container security platforms can scan image repositories and compare each against databases of known vulnerabilities and issues. Scans can be automatically triggered and scheduled, helping prevent the addition of harmful elements during development and in production. Auto-generated audit reports can be tracked against standard benchmarks, or an organisation can set its own security standards – useful in environments where highly-sensitive data is processed.

The connectivity between specialist container security functions and orchestration software means that suspected containers can be isolated or closed immediately, insecure permissions revoked, and user access suspended. With API connections to local firewalls and VPN endpoints, entire environments or subnets can be isolated, or traffic stopped at network borders.

Final word

Machine learning can reduce the risk of data breach in containerised environments by working on several levels. Anomaly detection, asset scanning, and flagging potential misconfiguration are all possible, plus any degree of automated alerting or amelioration are relatively simple to enact.

The transformative possibilities of container-based apps can be approached without the security issues that have stopped some from exploring, developing, and running microservice-based applications. The advantages of cloud-native technologies can be won without compromising existing security standards, even in high-risk sectors.

(Image source)

The post The role of machine learning in enhancing cloud-native container security appeared first on AI News.

The application of preemptive observability in pre-production may be more important as AI code generators become more common , the company claims. For instance, a 2023 Stanford University study revealed that developers using AI coding assistants were more likely to introduce bugs to their code. Despite this, major companies like Google are increasing their reliance on AI-generated code, with over 25% of the company’s new code being AI-created.

Nir Shafrir, CEO and Co-founder of Digma, commented on the growing resources that are being dedicated to ensuring systems perform well, saying, “We’re seeing a lot of effort invested in assuring optimal system performance, but many issues are still being discovered in complex code bases late in production.”

“Beyond this, scaling has often remained a rough estimation in organisations anticipating growth, and many are hitting barriers in technology growth that arise precisely during periods of significant organisational expansion. It means that engineering teams may spend between 20-40% of their time addressing issues discovered late in production environments, with some organisations spending up to 50% of engineering resources on fixing production problems.”

Preemptive observability is expected to become a key factor helping companies gain competitive advantage. It has several potential benefits for AI-generated code, including speed increases and improvements to the reliability of human-written code. According to Digma, preemptive observability helps ensure manually written code is more trustworthy, and reduces risk in the final product.

As well as tackling bugs introduced by AI code generation, Digma’s preemptive observability analysis engine has been designed to combat common, long-established issues companies may have experienced with human-made code, which may result in service level agreement (SLA) violations and performance issues. For high transactional establishments, like retail, fintech, and e-commerce, this technology could become valuable.

Digma’s algorithm has been designed to use pattern matching and anomaly detection techniques to analyse data and find specific behaviours or issues. It is capable of predicting what an application’s response times and resource usage should be, identifying possible issues before they can cause any noticeable damage. Digma specifically detects the part of the code that is causing an issue by analysing tracing data.

Preemptive observability analysis prevents problems rather than dealing with the aftermath of the issues. Teams can monitor holistically, and address potential issues in areas that are frequently ignored once in production.

Roni Dover, CTO and Co-founder of Digma, highlighted what differentiates Digma’s preemptive observability analysis engine from others: “By understanding runtime behaviour and suggesting fixes for performance issues, scaling problems, and team conflicts, we’re helping enterprises prevent problems and reduce risks proactively rather than putting out fires in production.”

Application performance monitoring (APM) tools are used to identify service issues, monitor production statuses, and highlight SLA errors. APMs are practical for sending alerts when services fail or slow during production. But unlike preemptive observability, APMs are limited in non-production settings, and can’t provide analysis of problems’ sources.

By identifying performance and scaling issues early on in the production process, even when data volumes are low, preemptive observability helps prevent major problems and reduce cloud costs.

Digma recently completed a successful $6 million seed funding round, indicating a growing confidence in the technology.

Image source: “Till Bechtolsheimer’s – Alfa Romeo Giulia Sprint GT No.40 – 2013 Donington Historic Festival” by Motorsport in Pictures is licensed under CC BY-NC-SA 2.0.

See also: Microsoft and OpenAI probe alleged data theft by DeepSeek

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post Digma’s preemptive observability engine cuts code issues, streamlines AI appeared first on AI News.

The statement was issued in response to a lawsuit filed by Indian news agency ANI, which accused the AI business of using its content without permission.

The Microsoft-backed AI giant stated that Indian courts lack jurisdiction in the case since OpenAI has no offices nor operations in the country. In its January 10 filing to the Delhi High Court, OpenAI emphasised that it is already defending similar lawsuits in the US, where it is required to preserve its training data during ongoing litigation.

The case, filed by ANI in November, is one of India’s most closely-watched lawsuits involving the use of AI. ANI alleges that OpenAI utilised its published content without authorisation to train ChatGPT and is demanding the deletion of its data from the company’s systems.

A global battle over copyright and AI

OpenAI is no stranger to such disputes, facing a wave of lawsuits from copyright holders worldwide. In the US, the New York Times filed a similar case against the company, accusing it of misusing its content. OpenAI has consistently denied such allegations, claiming its systems rely on the fair use of publicly available data.

During a November hearing in Delhi, OpenAI told the court it would no longer use ANI’s content. However, ANI argued that its previously published material remains stored in ChatGPT’s repositories and must be deleted.

In its rebuttal, OpenAI highlighted that it is legally obligated under US law to retain training data while related cases are pending. “The company is under a legal obligation, under the laws of the United States, to preserve, and not delete, the said training data,” OpenAI stated in its filing.

Jurisdiction dispute

OpenAI also argued that the relief ANI is seeking falls outside the jurisdiction of Indian courts. It pointed out that the company has “no office or permanent establishment in India,” and its servers, which store ChatGPT’s training data, are located outside the country.

ANI, which is partially owned by Reuters, countered the claim, saying the Delhi court has the authority to hear the case and that it will file a detailed response.

A Reuters spokesperson declined to comment on proceedings, but has stated that the agency has no involvement in ANI’s business operations.

Concerns over competition

ANI has also expressed concern about unfair competition, citing OpenAI’s partnerships with major news organisations like Time Magazine, The Financial Times, and France’s Le Monde. ANI says that these agreements give OpenAI an edge.

The agency further claimed that ChatGPT reproduces verbatim or similar excerpts of its works in response to user prompts. OpenAI, on the other hand, claimed that ANI deliberately used its own articles as prompts to “manipulate ChatGPT” to file the lawsuit.

The case is scheduled to be heard by the Delhi High Court on January 28. Meanwhile, OpenAI is transitioning from a non-profit to a for-profit company, raising $6.6 billion last year.

In recent months, OpenAI has secured high-profile deals with media outlets from around the world, highlighting its efforts to strengthen its commercial partnerships while managing regulatory concerns worldwide.

(Photo by Unsplash)

See also: DeepSeek-R1 reasoning models rival OpenAI in performance 

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post OpenAI argues against ChatGPT data deletion in Indian court appeared first on AI News.

AI News caught up with Nerijus Šveistys, Senior Legal Counsel at Oxylabs, to understand the state of play when it comes to AI regulation and its potential implications for industries, businesses, and innovation.

“The boom of the last few years appears to have sparked a push to establish regulatory frameworks for AI governance,” explains Šveistys.

“This is a natural development, as the rise of AI seems to pose issues in data privacy and protection, bias and discrimination, safety, intellectual property, and other legal areas, as well as ethics that need to be addressed.”

Regions diverge in regulatory strategy

The European Union’s AI Act has, unsurprisingly, positioned the region with a strict, centralised approach. The regulation, which came into force this year, is set to be fully effective by 2026.

Šveistys pointed out that the EU has acted relatively swiftly compared to other jurisdictions: “The main difference we can see is the comparative quickness with which the EU has released a uniform regulation to govern the use of all types of AI.”

Meanwhile, other regions have opted for more piecemeal approaches. China, for instance, has been implementing regulations specific to certain AI technologies in a phased-out manner. According to Šveistys, China began regulating AI models as early as 2021.

“In 2021, they introduced regulation on recommendation algorithms, which [had] increased their capabilities in digital advertising. It was followed by regulations on deep synthesis models or, in common terms, deepfakes and content generation in 2022,” he said.

“Then, in 2023, regulation on generative AI models was introduced as these models were making a splash in commercial usage.”

The US, in contrast, remains relatively uncoordinated in its approach. Federal-level regulations are yet to be enacted, with efforts mostly emerging at the state level.

“There are proposed regulations at the state level, such as the so-called California AI Act, but even if they come into power, it may still take some time before they do,” Šveistys noted.

This delay in implementing unified AI regulations in the US has raised questions about the extent to which business pushback may be contributing to the slow rollout. Šveistys said that while lobbyist pressure is a known factor, it’s not the only potential reason.

“There was pushback to the EU AI Act, too, which was nevertheless introduced. Thus, it is not clear whether the delay in the US is only due to lobbyism or other obstacles in the legislation enactment process,” explains Šveistys.

“It might also be because some still see AI as a futuristic concern, not fully appreciating the extent to which it is already a legal issue of today.”

Balancing innovation and safety

Differentiated regulatory approaches could affect the pace of innovation and business competitiveness across regions.

Europe’s regulatory framework, though more stringent, aims to ensure consumer protection and ethical adherence—something that less-regulated environments may lack.

“More rigid regulatory frameworks may impose compliance costs for businesses in the AI field and stifle competitiveness and innovation. On the other hand, they bring the benefits of protecting consumers and adhering to certain ethical norms,” comments Šveistys.

This trade-off is especially pronounced in AI-related sectors such as targeted advertising, where algorithmic bias is increasingly scrutinised.

AI governance often extends beyond laws that specifically target AI, incorporating related legal areas like those governing data collection and privacy. For example, the EU AI Act also regulates the use of AI in physical devices, such as elevators.

“Additionally, all businesses that collect data for advertisement are potentially affected as AI regulation can also cover algorithmic bias in targeted advertising,” emphasises Šveistys.

Impact on related industries

One industry that is deeply intertwined with AI developments is web scraping. Typically used for collecting publicly available data, web scraping is undergoing an AI-driven evolution.

“From data collection, validation, analysis, or overcoming anti-scraping measures, there is a lot of potential for AI to massively improve the efficiency, accuracy, and adaptability of web scraping operations,” said Šveistys. 

However, as AI regulation and related laws tighten, web scraping companies will face greater scrutiny.

“AI regulations may also bring the spotlight on certain areas of law that were always very relevant to the web scraping industry, such as privacy or copyright laws,” Šveistys added.

“At the end of the day, scraping content protected by such laws without proper authorisation could always lead to legal issues, and now so can using AI this way.”

Copyright battles and legal precedents

The implications of AI regulation are also playing out on a broader legal stage, particularly in cases involving generative AI tools.

High-profile lawsuits have been launched against AI giants like OpenAI and its primary backer, Microsoft, by authors, artists, and musicians who claim their copyrighted materials were used to train AI systems without proper permission.

“These cases are pivotal in determining the legal boundaries of using copyrighted material for AI development and establishing legal precedents for protecting intellectual property in the digital age,” said Šveistys.

While these lawsuits could take years to resolve, their outcomes may fundamentally shape the future of AI development. So, what can businesses do now as the regulatory and legal landscape continues to evolve?

“Speaking about the specific cases of using copyrighted material for AI training, businesses should approach this the same way as any web-scraping activity – that is, evaluate the specific data they wish to collect with the help of a legal expert in the field,” recommends Šveistys.

“It is important to recognise that the AI legal landscape is very new and rapidly evolving, with not many precedents in place to refer to as of yet. Hence, continuous monitoring and adaptation of your AI usage are crucial.”

Just this week, the UK Government made headlines with its announcement of a consultation on the use of copyrighted material for training AI models. Under the proposals, tech firms could be permitted to use copyrighted material unless owners have specifically opted out.

Despite the diversity of approaches globally, the AI regulatory push marks a significant moment for technological governance. Whether through the EU’s comprehensive model, China’s step-by-step strategy, or narrower, state-level initiatives like in the US, businesses worldwide must navigate a complex, evolving framework.

The challenge ahead will be striking the right balance between fostering innovation and mitigating risks, ensuring that AI remains a force for good while avoiding potential harms.

(Photo by Nathan Bingle)

See also: Anthropic urges AI regulation to avoid catastrophes

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post AI governance: Analysing emerging global regulations appeared first on AI News.

New records, part of the Algorithmic Transparency Recording Standard (ATRS), were published this week to shed light on the AI tools being used and set a benchmark for transparency and accountability in the integration of technology in public service delivery.

The initiative is part of the government’s broader strategy to embrace technology to improve outcomes, echoing commitments outlined in the “Plan for Change” to modernise public services and drive economic growth through innovative solutions.

The power of AI for modernisation

Among the published records, the Foreign, Commonwealth and Development Office is leveraging AI to provide faster responses to Britons seeking assistance overseas. Similarly, the Ministry of Justice is utilising algorithms to help researchers gain a deeper understanding of how individuals interact with the justice system, while other departments are deploying AI to enhance job advertisements.

The ATRS aims to document how such algorithmic tools are utilised and ensure their responsible application. By doing so, the government hopes to strengthen public trust in these innovations while encouraging their continued adoption across sectors.

Speaking on the government’s approach, Science Secretary Peter Kyle remarked:  

“Technology has huge potential to transform public services for the better; we will put it to use to cut backlogs, save money, and improve outcomes for citizens across the country.

Transparency in how and why the public sector is using algorithmic tools is crucial to ensure that they are trusted and effective. That is why we will continue to take bold steps like releasing these records to make sure everyone is clear on how we are applying and trialling technology as we use it to bring public services back from the brink.”

Specifically, the Department for Business and Trade has highlighted its algorithmic tool designed to predict which companies are likely to export goods internationally.

The AI-driven approach allows officials to target support towards high-growth potential businesses, enabling them to reach global markets faster. Previously reliant on time-consuming manual methods to analyse the more than five million companies registered on Companies House, this advancement ensures better allocation of resources and expedited assistance.

Business Secretary Jonathan Reynolds said:  

“Our Plan for Change will deliver economic growth, and for that to succeed, we need to support companies across the UK to realise their full potential when it comes to exporting around the globe.

Our use of AI plays a vital and growing role in that mission, allowing high-growth businesses to maximise the export opportunities available to them, while ensuring that we are using taxpayers’ money responsibly and efficiently in delivering economic stability.”

Establishing clear guidelines for AI in public services

To bolster public trust, new guidelines have been announced to clarify the scope of algorithmic transparency records.

Central government organisations will need to publish a record for any algorithmic tool that interacts directly with citizens or plays a significant role in decision-making about individuals. Limited exceptions, such as those concerning national security, apply.  

These records will be published once tools are piloted publicly or have become operational. They will detail the data used to train AI models, the underlying technologies, and the measures implemented to mitigate risks.

Importantly, the records also seek to confirm that – while AI tools are used to accelerate decision-making processes – human oversight remains integral, with trained staff responsible for final decisions.

Dr Antonio Espingardeiro, a member of IEEE and an expert in software and robotics, commented:

“AI has the potential to radically transform the public sector. In recent years, we have seen AI become a credible part of everyday public services. As it becomes more sophisticated, AI can conduct data-heavy tasks traditionally undertaken by humans. It can analyse vast quantities of information and, when coupled with machine learning, search through records and infer patterns or anomalies in data that would otherwise take decades for humans to analyse.

With this announcement, the UK government has acknowledged AI’s potential and proven that technology investment is essential to improving outcomes and the delivery of vital services. Over time, machine learning and generative AI (GenAI) could bring substantial value to the public system. With increased adoption, we will soon be able to deliver the scalability that the public sector needs and relieve the pressures and workloads placed on staff.”

Eleanor Watson, also a member of IEEE and an AI ethics engineer affiliated with Singularity University, added:

“With AI growing more rapidly than ever before, and already being tested and employed in education, healthcare, transportation, finance, data security, and more, the government, tech leaders, and academia should work together to establish standards and regulations for safe and responsible development of AI-based systems. This way, AI can be used to its full potential as indicated with this latest announcement.

Data privacy is probably the most critical ethical consideration, requiring informed consent, data anonymisation, strict access controls, secure storage, and compliance. New techniques such as homomorphic encryption, zero-knowledge proofs, federated learning, and part-trained models can help models to make use of our personal data in an encrypted form.”

Transparency remains a key tenet of the UK Government’s AI strategy. This announcement follows a recent statement by Pat McFadden, Chancellor of the Duchy of Lancaster, who affirmed that the benefits of technology – particularly AI – must span both public and private sectors and be used to modernise government.

As the Science Secretary’s department solidifies government efforts to create a “digital centre,” it marks a major step forward in boosting the responsible and effective use of AI across the UK’s public sector.

The ATRS records offer a valuable template for how governments worldwide can deploy AI systems to maximise efficiency, grow transparency, and balance the need for innovation with ethical considerations.

(Photo by Shreyas Sane)

See also: MHRA pilots ‘AI Airlock’ to accelerate healthcare adoption

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post UK wants to prove AI can modernise public services responsibly appeared first on AI News.

Announced on December 3, 2024, the public preview release lets organisations orchestrate AI workloads through a unified control plane, whether those workloads are running on cloud, on-premises, or in air-gapped infrastructure. The platform can work with any AI model and hardware accelerator including GPUs, CPUs, and TPUs.

“Clarifai has always been ahead of the curve, with over a decade of experience supporting large enterprise and mission-critical government needs with the full stack of AI tools to create custom AI workloads,” said Matt Zeiler, founder and CEO of Clarifai. “Now, we’re opening up capabilities we built internally to optimise our compute costs as we scale to serve millions of models simultaneously.”

The company claims its platform can reduce compute usage by 3.7x through model packing optimisations while supporting over 1.6 million inference requests per second with 99.9997% reliability. According to Clarifai, the optimisations can potentially cut costs by 60-90%, depending on configuration.

Capabilities of the compute orchestration platform include:

• Cost optimisation through automated resource management, including model packing, dependency simplification, and customisable auto-scaling options that can scale to zero for model replicas and compute nodes,
• Deployment flexibility on any hardware vendor including cloud, on-premise, air-gapped, and Clarifai SaaS infrastructure,
• Integration with Clarifai’s AI platform for data labeling, training, evaluation, workflows, and feedback,
• Security features that allow deployment into customer VPCs or on-premise Kubernetes clusters without requiring open inbound ports, VPC peering, or custom IAM roles.

The platform emerged from Clarifai customers’ issues with AI performance and cost. “If we had a way to think about it holistically and look at our on-prem costs compared to our cloud costs, and then be able to orchestrate across environments with a cost basis, that would be incredibly valuable,” noted a customer, as cited in Clarifai’s announcement.

The compute orchestration capabilities build on Clarifai’s existing AI platform that, the company says, has processed over 2 billion operations in computer vision, language, and audio AI. The company reports maintaining 99.99%+ uptime and 24/7 availability for critical applications.

The compute orchestration capability is currently available in public preview. Organisations interested in testing the platform should contact Clarifai for access.

The post New Clarifai tool orchestrates AI across any infrastructure appeared first on AI News.

A year after introducing the Gemini 1.0 model, this major upgrade incorporates enhanced multimodal capabilities, agentic functionality, and innovative user tools designed to push boundaries in AI-driven technology.

Leap towards transformational AI  

Reflecting on Google’s 26-year mission to organise and make the world’s information accessible, Pichai remarked, “If Gemini 1.0 was about organising and understanding information, Gemini 2.0 is about making it much more useful.”

Gemini 1.0, released in December 2022, was notable for being Google’s first natively multimodal AI model. The first iteration excelled at understanding and processing text, video, images, audio, and code. Its enhanced 1.5 version became widely embraced by developers for its long-context understanding, enabling applications such as the productivity-focused NotebookLM.

Now, with Gemini 2.0, Google aims to accelerate the role of AI as a universal assistant capable of native image and audio generation, better reasoning and planning, and real-world decision-making capabilities. In Pichai’s words, the development represents the dawn of an “agentic era.”

“We have been investing in developing more agentic models, meaning they can understand more about the world around you, think multiple steps ahead, and take action on your behalf, with your supervision,” Pichai explained.

Gemini 2.0: Core features and availability

At the heart of today’s announcement is the experimental release of Gemini 2.0 Flash, the flagship model of Gemini’s second generation. It builds upon the foundations laid by its predecessors while delivering faster response times and advanced performance.

Gemini 2.0 Flash supports multimodal inputs and outputs, including the ability to generate native images in conjunction with text and produce steerable text-to-speech multilingual audio. Additionally, users can benefit from native tool integration such as Google Search and even third-party user-defined functions.

Developers and businesses will gain access to Gemini 2.0 Flash via the Gemini API in Google AI Studio and Vertex AI, while larger model sizes are scheduled for broader release in January 2024.

For global accessibility, the Gemini app now features a chat-optimised version of the 2.0 Flash experimental model. Early adopters can experience this updated assistant on desktop and mobile, with a mobile app rollout imminent.

Products such as Google Search are also being enhanced with Gemini 2.0, unlocking the ability to handle complex queries like advanced math problems, coding enquiries, and multimodal questions.

Comprehensive suite of AI innovations  

The launch of Gemini 2.0 comes with compelling new tools that showcase its capabilities.

One such feature, Deep Research, functions as an AI research assistant, simplifying the process of investigating complex topics by compiling information into comprehensive reports. Another upgrade enhances Search with Gemini-enabled AI Overviews that tackle intricate, multi-step user queries.

The model was trained using Google’s sixth-generation Tensor Processing Units (TPUs), known as Trillium, which Pichai notes “powered 100% of Gemini 2.0 training and inference.”

Trillium is now available for external developers, allowing them to benefit from the same infrastructure that supports Google’s own advancements.

Pioneering agentic experiences  

Accompanying Gemini 2.0 are experimental “agentic” prototypes built to explore the future of human-AI collaboration, including:

• Project Astra: A universal AI assistant

First introduced at I/O earlier this year, Project Astra taps into Gemini 2.0’s multimodal understanding to improve real-world AI interactions. Trusted testers have trialled the assistant on Android, offering feedback that has helped refine its multilingual dialogue, memory retention, and integration with Google tools like Search, Lens, and Maps. Astra has also demonstrated near-human conversational latency, with further research underway for its application in wearable technology, such as prototype AI glasses.

• Project Mariner: Redefining web automation 

Project Mariner is an experimental web-browsing assistant that uses Gemini 2.0’s ability to reason across text, images, and interactive elements like forms within a browser. In initial tests, it achieved an 83.5% success rate on the WebVoyager benchmark for completing end-to-end web tasks. Early testers using a Chrome extension are helping to refine Mariner’s capabilities while Google evaluates safety measures that ensure the technology remains user-friendly and secure.

• Jules: A coding agent for developers  

Jules, an AI-powered assistant built for developers, integrates directly into GitHub workflows to address coding challenges. It can autonomously propose solutions, generate plans, and execute code-based tasks—all under human supervision. This experimental endeavour is part of Google’s long-term goal to create versatile AI agents across various domains.

• Gaming applications and beyond  

Extending Gemini 2.0’s reach into virtual environments, Google DeepMind is working with gaming partners like Supercell on intelligent game agents. These experimental AI companions can interpret game actions in real-time, suggest strategies, and even access broader knowledge via Search. Research is also being conducted into how Gemini 2.0’s spatial reasoning could support robotics, opening doors for physical-world applications in the future.

Addressing responsibility in AI development

As AI capabilities expand, Google emphasises the importance of prioritising safety and ethical considerations.

Google claims Gemini 2.0 underwent extensive risk assessments, bolstered by the Responsibility and Safety Committee’s oversight to mitigate potential risks. Additionally, its embedded reasoning abilities allow for advanced “red-teaming,” enabling developers to evaluate security scenarios and optimise safety measures at scale.

Google is also exploring safeguards to address user privacy, prevent misuse, and ensure AI agents remain reliable. For instance, Project Mariner is designed to prioritise user instructions while resisting malicious prompt injections, preventing threats like phishing or fraudulent transactions. Meanwhile, privacy controls in Project Astra make it easy for users to manage session data and deletion preferences.

Pichai reaffirmed the company’s commitment to responsible development, stating, “We firmly believe that the only way to build AI is to be responsible from the start.”

With the Gemini 2.0 Flash release, Google is edging closer to its vision of building a universal assistant capable of transforming interactions across domains.

See also: Machine unlearning: Researchers make AI models ‘forget’ data

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post Gemini 2.0: Google ushers in the agentic AI era  appeared first on AI News.

Progress in AI has provided tools capable of revolutionising various domains, from healthcare to autonomous driving. However, as technology advances, so do its complexities and ethical considerations. 

The paradigm of large-scale pre-trained AI systems, such as OpenAI’s ChatGPT and CLIP (Contrastive Language–Image Pre-training), has reshaped expectations for machines. These highly generalist models, capable of handling a vast array of tasks with consistent precision, have seen widespread adoption for both professional and personal use.  

However, such versatility comes at a hefty price. Training and running these models demands prodigious amounts of energy and time, raising sustainability concerns, as well as requiring cutting-edge hardware significantly more expensive than standard computers. Compounding these issues is that generalist tendencies may hinder the efficiency of AI models when applied to specific tasks.  

For instance, “in practical applications, the classification of all kinds of object classes is rarely required,” explains Associate Professor Go Irie, who led the research. “For example, in an autonomous driving system, it would be sufficient to recognise limited classes of objects such as cars, pedestrians, and traffic signs.

“We would not need to recognise food, furniture, or animal species. Retaining classes that do not need to be recognised may decrease overall classification accuracy, as well as cause operational disadvantages such as the waste of computational resources and the risk of information leakage.”  

A potential solution lies in training models to “forget” redundant or unnecessary information—streamlining their processes to focus solely on what is required. While some existing methods already cater to this need, they tend to assume a “white-box” approach where users have access to a model’s internal architecture and parameters. Oftentimes, however, users get no such visibility.  

“Black-box” AI systems, more common due to commercial and ethical restrictions, conceal their inner mechanisms, rendering traditional forgetting techniques impractical. To address this gap, the research team turned to derivative-free optimisation—an approach that sidesteps reliance on the inaccessible internal workings of a model.  

Advancing through forgetting

The study, set to be presented at the Neural Information Processing Systems (NeurIPS) conference in 2024, introduces a methodology dubbed “black-box forgetting.”

The process modifies the input prompts (text instructions fed to models) in iterative rounds to make the AI progressively “forget” certain classes. Associate Professor Irie collaborated on the work with co-authors Yusuke Kuwana and Yuta Goto (both from TUS), alongside Dr Takashi Shibata from NEC Corporation.  

For their experiments, the researchers targeted CLIP, a vision-language model with image classification abilities. The method they developed is built upon the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), an evolutionary algorithm designed to optimise solutions step-by-step. In this study, CMA-ES was harnessed to evaluate and hone prompts provided to CLIP, ultimately suppressing its ability to classify specific image categories.

As the project progressed, challenges arose. Existing optimisation techniques struggled to scale up for larger volumes of targeted categories, leading the team to devise a novel parametrisation strategy known as “latent context sharing.”  

This approach breaks latent context – a representation of information generated by prompts – into smaller, more manageable pieces. By allocating certain elements to a single token (word or character) while reusing others across multiple tokens, they dramatically reduced the problem’s complexity. Crucially, this made the process computationally tractable even for extensive forgetting applications.  

Through benchmark tests on multiple image classification datasets, the researchers validated the efficacy of black-box forgetting—achieving the goal of making CLIP “forget” approximately 40% of target classes without direct access to the AI model’s internal architecture.

This research marks the first successful attempt to induce selective forgetting in a black-box vision-language model, demonstrating promising results.  

Benefits of helping AI models forget data

Beyond its technical ingenuity, this innovation holds significant potential for real-world applications where task-specific precision is paramount.

Simplifying models for specialised tasks could make them faster, more resource-efficient, and capable of running on less powerful devices—hastening the adoption of AI in areas previously deemed unfeasible.  

Another key use lies in image generation, where forgetting entire categories of visual context could prevent models from inadvertently creating undesirable or harmful content, be it offensive material or misinformation.  

Perhaps most importantly, this method addresses one of AI’s greatest ethical quandaries: privacy.

AI models, particularly large-scale ones, are often trained on massive datasets that may inadvertently contain sensitive or outdated information. Requests to remove such data—especially in light of laws advocating for the “Right to be Forgotten”—pose significant challenges.

Retraining entire models to exclude problematic data is costly and time-intensive, yet the risks of leaving it unaddressed can have far-reaching consequences.

“Retraining a large-scale model consumes enormous amounts of energy,” notes Associate Professor Irie. “‘Selective forgetting,’ or so-called machine unlearning, may provide an efficient solution to this problem.”  

These privacy-focused applications are especially relevant in high-stakes industries like healthcare and finance, where sensitive data is central to operations.  

As the global race to advance AI accelerates, the Tokyo University of Science’s black-box forgetting approach charts an important path forward—not only by making the technology more adaptable and efficient but also by adding significant safeguards for users.  

While the potential for misuse remains, methods like selective forgetting demonstrate that researchers are proactively addressing both ethical and practical challenges.  

See also: Why QwQ-32B-Preview is the reasoning AI to watch

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post Machine unlearning: Researchers make AI models ‘forget’ data appeared first on AI News.

Reportedly led by a dozen AI researchers, scientists, and investors, the new training techniques, which underpin OpenAI’s recent ‘o1’ model (formerly Q* and Strawberry), have the potential to transform the landscape of AI development. The reported advances may influence the types or quantities of resources AI companies need continuously, including specialised hardware and energy to aid the development of AI models.

The o1 model is designed to approach problems in a way that mimics human reasoning and thinking, breaking down numerous tasks into steps. The model also utilises specialised data and feedback provided by experts in the AI industry to enhance its performance.

Since ChatGPT was unveiled by OpenAI in 2022, there has been a surge in AI innovation, and many technology companies claim existing AI models require expansion, be it through greater quantities of data or improved computing resources. Only then can AI models consistently improve.

Now, AI experts have reported limitations in scaling up AI models. The 2010s were a revolutionary period for scaling, but Ilya Sutskever, co-founder of AI labs Safe Superintelligence (SSI) and OpenAI, says that the training of AI models, particularly in the understanding language structures and patterns, has levelled off.

“The 2010s were the age of scaling, now we’re back in the age of wonder and discovery once again. Scaling the right thing matters more now,” they said.

In recent times, AI lab researchers have experienced delays in and challenges to developing and releasing large language models (LLM) that are more powerful than OpenAI’s GPT-4 model.

First, there is the cost of training large models, often running into tens of millions of dollars. And, due to complications that arise, like hardware failing due to system complexity, a final analysis of how these models run can take months.

In addition to these challenges, training runs require substantial amounts of energy, often resulting in power shortages that can disrupt processes and impact the wider electriciy grid. Another issue is the colossal amount of data large language models use, so much so that AI models have reportedly used up all accessible data worldwide.

Researchers are exploring a technique known as ‘test-time compute’ to improve current AI models when being trained or during inference phases. The method can involve the generation of multiple answers in real-time to decide on a range of best solutions. Therefore, the model can allocate greater processing resources to difficult tasks that require human-like decision-making and reasoning. The aim – to make the model more accurate and capable.

Noam Brown, a researcher at OpenAI who helped develop the o1 model, shared an example of how a new approach can achieve surprising results. At the TED AI conference in San Francisco last month, Brown explained that “having a bot think for just 20 seconds in a hand of poker got the same boosting performance as scaling up the model by 100,000x and training it for 100,000 times longer.”

Rather than simply increasing the model size and training time, this can change how AI models process information and lead to more powerful, efficient systems.

It is reported that other AI labs have been developing versions of the o1 technique. The include xAI, Google DeepMind, and Anthropic. Competition in the AI world is nothing new, but we could see a significant impact on the AI hardware market as a result of new techniques. Companies like Nvidia, which currently dominates the supply of AI chips due to the high demand for their products, may be particularly affected by updated AI training techniques.

Nvidia became the world’s most valuable company in October, and its rise in fortunes can be largely attributed to its chips’ use in AI arrays. New techniques may impact Nvidia’s market position, forcing the company to adapt its products to meet the evolving AI hardware demand. Potentially, this could open more avenues for new competitors in the inference market.

A new age of AI development may be on the horizon, driven by evolving hardware demands and more efficient training methods such as those deployed in the o1 model. The future of both AI models and the companies behind them could be reshaped, unlocking unprecedented possibilities and greater competition.

See also: Anthropic urges AI regulation to avoid catastrophes

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, a

The post New AI training techniques aim to overcome current challenges appeared first on AI News.