AI Weekly Digest - July 1 2024

A selection of fascinating stories from the speedily progressing AI sector. LinkedIn newsletter readers can also sign up to an enhanced email edition of the AI Weekly Digest - published every Friday. You can subscribe for free.

Bill Gates: 'AI will accelerate innovation, combat climate change'

Bill Gates said: "AI, in every field of endeavour, will be accelerating innovation." Interviewed by Sky News at the Breakthrough Energy Summit in London, the Microsoft co-founder said the technology must be "used by people with good intent". He said AI would make it easier to combat climate change, adding: "Thank goodness AI is going to make that?far easier to do." Asked if the technology could potentially be used to overthrow governments, the philanthropist said he had "not heard that particular scenario", but added: "AI could be used by" people planning cyber-attacks or political interference. Gates concluded: "So you have to make sure the good guys are staying ahead in detecting and preventing that type of usage."

Reports: Meta and Apple held talks over AI link-up

Meta has reportedly held talks on integrating its generative AI model into Apple's upcoming AI system for iPhones. The Wall Street Journal reported on the possible tie-up, which followed Apple Intelligence signing a deal with OpenAI’s ChatGPT and saying it also wanted to integrate Google’s Gemini. The reports said discussions between the Facebook and Instagram owner and Apple have not been finalised and could still fall through.?Meanwhile, Bloomberg later reported it understands Apple has shelved the idea of integrating Meta's AI models with iPhones due to privacy concerns.

Suleyman: 'Stop framing AI as a ferocious race'

Microsoft?AI chief executive Mustafa Suleyman said "we have to stop framing (artificial intelligence) as a ferocious race, telling the Aspen Ideas Festival: “I don’t buy the metaphor that there is a finish line. This is another false frame.” He also insisted Microsoft is in "ferocious competition" with?OpenAI, despite investing $10bn in the group. He said: "They are an independent company. We don’t own or control them. We don’t even have any board members. So, they do entirely their own thing. But we have a deep partnership. I’m very good friends with (CEO) Sam (Altman), have huge respect and, trust and faith in what they’ve done. And that’s how it’s going to roll for many, many years to come."

Claim: AI groups bypass protocol to stop publishers' content being scraped

Content licensing start-up TollBit has claimed artificial intelligence groups are bypassing robots.txt - the protocol deployed by publishers to prevent their content being scraped. The US?News/Media Alliance?said: "Without the ability to opt out of massive scraping, we cannot monetise our valuable content and pay journalists. This could seriously harm our industry."

YouTube talks with record labels about AI licensing deals

YouTube has opened talks with record labels about AI licensing deals, seeking permission to use popular artists' catalogues to train song generators. The group has reportedly offered?upfront cash payments to the three majors - Sony, Warner and Universal - despite many artists being opposed to automated music generation. YouTube said: "We are always testing new ideas and learning from our experiments; it’s an important part of our innovation process. We will continue on this path with AI and music as we build for the future.”

Union agrees AI-use deal with Hollywood studios

Hollywood workers union the International Alliance of Theatrical Stage Employees (IATSE) has signed an AI-use deal with leading?studios. The agreement is with the Alliance of Motion Picture and Television Producers (AMPTP), which represents?Netflix?and?Disney, among others. The deal, which also sets out pay increases, includes "language that ensures no employee is required to provide Al prompts in any manner that would result in the displacement of any covered employee".

Reddit updates Robots Exclusion Protocol

Reddit?said it has updated its Robots Exclusion Protocol to ensure AI groups do not scrape its content without permission to train large language models. It said: “Anyone accessing Reddit content must abide by our policies, including those in place to protect redditors. We are selective about who we work with and trust with large-scale access to Reddit content.”

Music labels sue Suno and Udio over 'mass copyright infringements'

Leading record labels?Sony Music, Universal Music Group and Warner Records have sued AI groups Suno and Udio in the US, alleging mass copyright infringements by using their recordings to train music-generating AI systems. Federal lawsuits filed in New York and Massachusetts accused Udio and Sudo of copying music, without permission, to teach their systems to create music which will "directly compete with, cheapen, and ultimately drown out" human artists' work.

Time latest to sign OpenAI content deal

Time magazine has signed a multi-year content deal with OpenAI for undisclosed terms, giving the ChatGPT owner access to?the US title's?vast archive of news content. The companies said that the?chatbot will cite and link back to the source on Time.com, in response to user queries. OpenAI's Time deal adds to similar ones in recent months with the Financial Times, Axel Springer, Le Monde and Spain's Prisa Media.

Open AI 'excited' after Rockset acquisition

OpenAI has made its first acquisition, buying real-time search and data analytics tools provider Rockset for undisclosed terms. COO Brad Lightcap said: “Rockset’s infrastructure empowers companies to transform their data into actionable intelligence. We’re excited to bring these benefits to our customers by integrating Rockset’s foundation into OpenAI products.”

Anthropic's cornerstone is 'trustworthy’ AI

Anthropic head Daniela Amodei has insisted that the cornerstone of the business is to produce "trustworthy’" AI. The Times quoted Amodei stating: “We take the responsibility of what we’re building very seriously. We’re a company that was founded with a goal of building this technology in a way that is very useful for people and that is reliable and trustworthy and safe”.

SoftBank to invest in Perplexity AI as value soars to $3bn

SoftBank Group's Vision Fund 2 will reportedly invest between $10m and $20m in US search startup Perplexity AI at a valuation of $3bn. Bloomberg News said the Japanese tech investor's stake would be part of its larger $250m funding round. SoftBank and Perplexity declined to comment. In January Perplexity AI raised $73.6m from a group of investors, including Nvidia, and Amazon founder Jeff Bezos. It was then valued at $520m.

Reports: ByteDance working with?Broadcom on advanced processor

TikTok owner ByteDance is reportedly working with US chip designer Broadcom to create an advanced AI processor. Two sources spoke about the tie-up, which Reuters suggested would help the Chinese group 'secure sufficient supply of high-end chips amid US-Sino tensions'. The report said the 5-nanometre application-specific integrated chip (ASIC) would be compliant with US export restrictions, with manufacturing work outsourced to Taiwan's?TSMC. Another source said that although development work is well under way, manufacturing is not expected to begin until next year. TSMC declined to comment.

Amazon set to charge for AI-powered Remarkable Alexa?

Amazon is reportedly planning to launch a paid-for version of its loss-making Alexa platform, charging $5-$10 per month for the generative AI-powered Remarkable Alexa. It would be the first overhaul since the voice assistant was launched in 2014.

Next government 'must create system to log AI-related incidents'

Thinktank the Centre for Long-Term Resilience has urged the next government to establish a system for logging AI-related incidents across the UK. Policy manager Tommy Shaffer Shane said: “Incident reporting has played a transformative role in mitigating and managing risks in safety-critical industries such as aviation and medicine. But it’s largely missing from the regulatory landscape being developed for AI. This is leaving the UK government blind to the incidents that are emerging from AI’s use, inhibiting its ability to respond."

EU vote scrapped on plan to use AI to uncover abuse material

An EU vote on proposals to use AI to scan social media accounts to root out child abuse material was scrapped after member states were unable to agree a deal. A spokesman for the Belgian?EU presidency said: "In the last hours before the meeting it… was clear that the required qualified majority could just not be met so we decided to remove the item from the agenda and to continue the consultation between the Member States - to continue working on the text.”

Centre for Investigative Reporting sues OpenAI

The Centre for Investigative Reporting has become the latest US news group to sue OpenAI, alleging the ChatGPT developer “copied, used, abridged, and displayed CIR’s valuable content without CIR’s permission or authorisation, and without any compensation”. The non-profit is also suing key investor Microsoft.

Google's Gemini to be rolled out to students worldwide

Google?announced it is to make its Gemini chatbot available to teenage students in more than 100 countries, after previously restricting it to over-18s.

Deepfakes are 'most common misuse' of gen AI tools

A study by Google’s DeepMind concluded AI-generated deepfakes, which impersonate celebrities and politicians, are the most common misuse of generative AI tools, ahead of the production of misinformation to post online.

Start-up planning AI chip to take on?Nvidia

Etched, a two-year-old start-up established by Harvard dropouts, has raised $120m to develop a new AI chip to take on?Nvidia. Co-founder and CEO Gavin Uberti said: “We’re making the biggest bet in AI. If transformers go away, we’ll die. But if they stick around, we’re the biggest company of all time.”

Safety concerns bring delay to ChatGPT voice assistant feature

OpenAI is delaying the release of its voice assistant feature for ChatGPT to address safety concerns. It said: “We’re improving the model’s ability to detect and refuse certain content. We’re also working on improving the user experience and preparing our infrastructure to scale to millions while maintaining real-time responses.”