What in the AI is going on… April to June 2024

April 2024

One step closer to superhuman AI

Meta launched its newest generation of its open-source LLM (large language model), Llama 3, which is the technology that powers its AI systems. The launch of Llama 3 signals a step closer to human-like AI that can reason and bestow a memory. OpenAI also discussed plans for its next AI model, GPT-5. Like Llama 3, GPT-5 will see improvements to the model’s planning and reasoning capabilities, going beyond its current capability to deal with discrete tasks. GPT-5 is expected to be released later in the year.

Lawrence gets the job done

Lawhive, a London-based legal tech start-up received £9.5m investment from Google. Lawhive uses Lawrence, its AI system, to facilitate the provision of legal advice by automating administrative tasks that are usually completed by a paralegal or junior lawyer. Although Lawrence has already passed the Solicitors Qualifying Exam with flying colours, Lawhive plans to expand its software development team and continue to improve Lawrence’s capabilities. 

Free advice for British AI

The UK’s Digital Regulation Cooperation Forum (DCRF), comprising the ICO, CMA, Ofcom and the FCA, launched its new AI and Digital Hub. The Hub will provide free and informal advice to UK businesses developing new AI products or services that benefit consumers, businesses and/or the UK economy, and which come within scope of at least two of the four DCRF regulators. The aim of the hub is to promote innovation and support new technologies to come to market. See our full Summer 2024 Snapshot on the development.

Don’t scrape from Reddit

Reddit issued a strong warning that it will bring legal action against AI providers that are found using data from its platform without permission. The warning comes in the context of growing concerns around data scraping by AI companies and challenges regarding user consent. Reddit is ready and willing to agree to license its content to AI firms as evidenced by the Google-Reddit deal in February 2024 (See the Spring 2024 addition here), but will not tolerate unauthorised use of its platform. A win for user privacy or a win for Reddit’s bottom line.

May 2024

Wayve-ing the flag for Britain 

Wayve, a British AI company developing self-driving vehicles, received $1 billion of investment from SoftBank Group, Nvidia and Microsoft. Wayve will use the funding to develop its “embodied AI” technology which combines machine learning, robotics and computer vision to allow the AI solution to be used in a physical context. Wayve will be applying its embodied AI to self-driving vehicles for the very first time, with the aim of creating safer self-driving vehicles that are able to properly interact within is surroundings and navigation routes. 

Sony says no!

Sony Music became the latest company to warn AI firms to stop using its data without permission. Sony Music published a “Declaration of AI Training Opt Out” stating it “expressly prohibits and opts out of any text or data mining, web scraping or similar reproductions, extractions or uses (“TDM”) of any SME and/or SMP content…for any purposes, including in relation to training, developing or commercializing any AI system”. The intention of the declaration was to protect the rights of songwriters and recording artists. 

ICO wants to know more about Microsoft Recall 

The ICO raised privacy concerns over Microsoft’s forthcoming feature “Recall”. Recall, an optional feature, will take screenshots of a user’s screen every five seconds and will store the screenshot on-device to create a bank of information of which users can then search using natural language. Whilst Microsoft has stated Recall was built with a privacy-by-design ethos, the ICO wanted to find out more about the product to ensure appropriate safeguards are in place to protect user privacy. 

Seoul Searching for AI companies

Following the first international AI Safety Summit held in November 2023 in Bletchley Park (see our previous Snapshot), the UK Government co-hosted the AI Seoul Summit with the Republic of Korea. At the Summit, 16 companies including Google, Meta, Amazon, OpenAI and Samsung, agreed to follow voluntary safety commitments when developing frontier AI to ensure potential risks of the system are properly mitigated, before new technology is deployed. 

June 2024

Bigger is better

AI start-up, xAI, owned by Elon Musk, Nvidia, Supermicro and Dell have partnered up to work on a project to build the world’s largest supercomputer. The project will see the tech heavyweights build their own infrastructure including AI factories, liquid coolers and servers. Musk aims to have the supercomputer up and running in the second half of 2025.

Adobe’s user data commitment

Following concerns relating to the roll-out of updated Terms of Use, Adobe issued a statement to clarify how it uses user data. Adobe confirmed that user data remains under ownership of the user, user content loaded to its platforms is only scanned and reviewed to detect illegal content, and user data is not used to train its generative AI tools.

Ready to deliver

Waabi, an AI company based in Canada, received $200 million from a funding round which included Uber, Nvidia, Porsche and Volvo. The funding will be put towards achieving Waabi’s goal to have self-driving trucks powered by generative AI on the roads of Texas by 2025. Like Wayve, Waabi is developing the capability of AI systems to perform in a physical context.

Summer 2024