AI Weekly Digest - November 18 2024
Foxconn - global demand for AI servers 'insane'
Apple assembler Foxconn said demand for AI servers is currently “insane" as it posted third-quarter net profit up 14% to $1.52bn, on revenue up 20%. Chairman Young Liu said AI server sales were up 200% over the first nine months, and were expected to be the group's key growth driver in 2025.
Meta and OpenAI warn of increasing difficulty to evaluate AI models
Meta AI lead Ahmad Al-Dahle said "it is becoming more and more difficult to evaluate" artificial intelligence models, as the FT explored how developers are moving towards internal benchmarks in the absence of public tests. OpenAI's Mark Chen, meanwhile, said: "We are getting to the era where a lot of the human-written tests are no longer sufficient as a good barometer for how capable the models are. That creates a new challenge for us as a research world.”
BBC, C4, ITN, see appetite for pan-industry approach to AI
The BBC, Channel 4 and ITN indicated there is an increasing appetite for a pan-industry approach to AI as they set out ways in which they are exploring the emerging technology. Speaking at the AI Creative Summit, ITN director of technology, production and innovation Jon Roberts said: “Last year the conversations [around AI] were informal and almost secretive, when there were various working groups set up. It’s now becoming increasingly formal and there’s a transition to organised groups trying to form clear policy objectives." Channel 4 director of technology Grace Boswood added: “It’s really important we all have a common purpose, but it is complicated. We all have to live our own values and work out how we come together in order to implement policy.”
Nvidia and SoftBank create world's first AI/5G telecoms network
AI-focused chipmaker Nvidia and SoftBank Corp have created the world's first artificial intelligence and 5G telecoms network, a joint statement announced. The telecoms arm of SoftBank and the US chip group said the new network can run AI and 5G workloads concurrently, as an artificial intelligence radio access network (AI-RAN). SoftBank Group CEO Masayoshi Son, speaking alongside Nvidia CEO Jensen Huang, said: "Every other telco will have to follow this new wave".
US judge rejects news groups' copyright case against OpenAI
A New York judge has thrown out a copyright case against OpenAI brought by news groups AlterNet and Raw Story, one of the first in the country which accused the ChatGPT developer of illegally scraping articles. The judge said: "While plaintiffs provide third-party statistics indicating that an earlier version of ChatGPT generated responses containing significant amounts of plagiarised content, plaintiffs have not plausibly alleged that there is a ‘substantial risk’ that the current version of ChatGPT will generate a response plagiarising one of plaintiffs’ articles.”
WSJ testing AI-generated article summaries
The Wall Street Journal has begun testing AI-generated article summaries. Head of digital Taneth Evans said: “We are always assessing new technologies and methods of storytelling to provide more value to our subscribers. To that end, we are currently running a series of A/B tests to understand our users’ needs with regards to summarization. The newsroom does this hand-in-hand with colleagues in technology and while speaking with readers at every step of the way. We also disclose how we leverage artificial intelligence tools to support our journalism whenever it’s used.”
Particle sees 'high referral click-through' rates to publishers
Particle co-founder Sara Beykpour said the US AI news start-up, which aims to establish a mutually beneficial relationship with publishers, has achieved high referral click-through rates in beta tests. The app offers modes including 'Explain Like I’m Five' and 'Opposite Sides' for news searches, and Beykpour said: “We have a processing pipeline that takes related content and summarises it into bullet points, into a headline, sub-headline, and does all the extractions. Then... we pull out quotes and links and all sorts of relevant information about [the story]. And we have our own algorithms to rank, so that the most important or relevant link is the one that you see first - or what we think is the most important or relevant quote is the one that you see first.”
BBC AI training model rewrites copy before going to human sub-editors
Danijela Horak, head of AI research at the BBC, said the broadcaster has created a training model to rewrite copy that would then go to a human sub-editor before being published. She said: “We are very happy with the initial results and hopefully it will be deployed in Q1 next year." Horak added that research into deepfake detection by AI is “a big focus this year”.
China's Baidu aims to compete with OpenAI after AI applications rollout
Chinese search giant Baidu unveiled a range of new applications for its artificial intelligence technology. These include a text-to-image generator and a tool which allows users without coding expertise to develop new software applications. The group's switch to commercialisation of large language model (LLM) applications follows nearly two years of heavy investment in research and development, reports said, with Baidu seeking to be a viable alternative to OpenAI's GPT. Meanwhile, Baidu is reportedly planning to unveil a pair of glasses with a built-in AI assistant which would rival Meta's Ray-Bans.
Trade Desk chief: Kokai to be fully adopted by brands next year
Demand-side platform The Trade Desk's CEO Jeff Green told analysts the company expects its AI-fuelled digital advertising platform, Kokai, to be fully adopted by brands the end of next year.
Reports: X testing Grok with free users
X has reportedly begun testing Grok with free users, after previously restricting the AI chatbot to its premium tier.
US and allies urged to create North American Compact for AI
OpenAI has urged the US and its allies to establish the 'North American Compact for AI', a formal alliance to support AI infrastructure so they can better compete with China. The ChatGPT developer said: “AI presents an unmissable opportunity to reindustrialise the US, and through that, generate the kind of broad-based economic growth that will revitalise the American Dream. It also presents a national security imperative to protect our nation and our allies against a surging China by offering an AI shaped by democratic values, promoting individual choice and benefiting the most people possible.”
Kyle: UK 2025 legislation to safeguard against risks of AI
Technology secretary Peter Kyle said the UK will introduce legislation to safeguard against the risks of artificial intelligence in the next year. He also said the UK’s AI Safety Institute will be turned into an arms-length government body with “the independence to act fully in the interests of British citizens”.
Report: Microsoft set to mirror Apple by branding AI features 'Windows Intelligence'
Microsoft could be set to mirror rival Apple and brand its AI features as 'Windows Intelligence', unconfirmed social media reports have suggested. These followed references to Windows Intelligence in a new group policy document from the US technology giant.
Flawless launches tools to apply AI to film editing
Start-up Flawless AI hailed its new tools to apply artificial intelligence to film editing. The software reportedly allows filmmakers to adjust and edit their work, allowing actors’ lip movements to match scripts in multiple languages. Original dialogue can also be changed without the need for reshoots and can deliver “perfectly lip-synced visuals”, co-CEO Scott Mann said. The UK-born, California-based tech entrepreneur said: "Filmmaking is so monumentally cumbersome and expensive. The average film takes 871 days. I’m going to try and make a film in 87 days next year [using Flawless].”
Virgin Media O2 creates 'scambaiter' tool
Virgin Media O2 has created an AI-powered 'scambaiter' tool which poses as an elderly grandmother to tie-up phone scammers on phone calls. Director of fraud Murray MacKenzie said: "The newest member of our fraud-prevention team, Daisy, is turning the tables on scammers - outsmarting and outmanoeuvring them at their own cruel game simply by keeping them on the line."
Cisco CEO: On track to beat target of $1bn AI orders
Cisco CEO Chuck Robbins has flagged up its growth in artificial intelligence infrastructure: "We have earned more design wins and remain confident that we will exceed our target of $1bn of AI orders this fiscal year from web-scale customers."