The SEC Newgate AI Weekly
With a growing geopolitical artificial intelligence arms race afoot, moves by Nvidia to consolidate its monopoly on AI hardware and further consideration from the UK on how to position itself on the technology, it’s never a dull week in the world of AI. See the latest updates below…
TikTok working on generative AI disclosure tools
TikTok is working on a way for creators to disclose if their posts contain AI-generated content, representing a significant step towards tackling deepfake/misinformation on its platform, despite being in trial stages currently. Using the tool, creators must disclose AI-generated content, or risk having their content taken down.
Leading commentators such as Henry Ajder have raised doubts about how this will work in practice, including how the labelling approach puts more onus on creators than the platform, and given that social media is already saturated with synthetic and partially AI-altered content. Nevertheless, this could serve as the start of a blueprint for other tech and social media platforms to follow in the fight against a proliferation of misleading synthetic content produced by generative AI.
Nvidia announces an even more powerful chip
Nvidia’s latest chips, the A100 and the H100, have already taken the company to extraordinary heights as a result of the company’s near monopoly of AI tech, but the semiconductor maker is already focused on its next generation chip. Last week, it announced that it will release an even more powerful chip to run its large AI models in Q1 of next year. Its next super chip will have triple the memory capacity, allowing for far greater performance and productivity.
OpenAI release web crawling tool “GPTBot” to train models with latest information
OpenAI quietly released a web crawling tool last week with the aim of enhancing model accuracy and expanded capability. The tool is aimed at gathering publicly available data whilst avoiding sources that involve paywalls, personal data collection and those that host content that clashes with OpenAI’s policies. It aims to address one of the biggest complaints lobbied at ChatGPT, which is criticised for its lack of up-to-date information from the web in its models and its chatbot. OpenAI announced that websites could opt out of including their content for GPTBot’s AI model training, prompting significant websites such as the Verge immediately to state that they would be blocking the tool’s scraping.
This move will represent a tightrope operation from OpenAI in order to train its Large Language Models (LLMs) with up-to-date information, whilst surviving controversy that it is using public content fairly, and within privacy regulations.
Saudi Arabia and UAE push to secure chips to secure AI ambitions
Saudi Arabia and the United Arab Emirates are reported to be pushing to buy thousands of the high-performance Nvidia chips critical for building AI software, adding to the already red-hot interest in securing the chips that enable generative AI. According to reports, Saudi Arabia has bought at least 3,000 chips, whilst the UAE has already secured access to thousands of chips and has developed its own open-source large language model known as Falcon. People close to Nvidia have said that the company will ship around 550,000 of its latest chips, the H100 GPU, in 2023, primarily to US tech companies at its current pricing of around $40,000 per chip.
UK consider following US by restricting investment into Chinese AI tech
The UK Government has stated that it was “closely” considering implementing similar measures to those proposed by President Joe Biden aimed at restricting investment in the Chinese tech sector, with a focus on AI tech. The US president last week signed an executive order which gave the Treasury secretary powers to restrict US investment into Chinese firms operating in the semiconductor, microelectronic, quantum and AI sectors for national security purposes. In response to the USA’s intentions to clamp down on AI technology, it was also reported that China’s internet giants are rushing to secure $5 billion worth of Nvidia chips to build generative AI systems.
UK government cash support for AI in healthcare
The UK Government confirmed that it would allocate £13m investment into advancing research in AI within healthcare. The funding will support 22 projects across universities and NHS trusts within the UK, with focuses ranging from semi-autonomous surgical robotics for tumour removal to predictive health analysis based on existing conditions.
Planning for the UK’s global summit on AI safety
On the same day as the additional funding for AI research in healthcare, the UK Government also announced that tech expert Matt Clifford and former senior diplomat Jonathon Black will lead the preparations for its global summit on AI safety, an event that the UK Government sees as a kick-off for its ambitions to become the ‘geographical home’ of global AI safety. As Politico notes, the event is currently behind on planning stages, without a date and an invitation list.