Companies and countries are being drawn into a dangerous race to adopt foundational artificial intelligence models like GPT-4. Calls for a pause on the race have been met with criticism, but could see the safety-conscious fall behind while others surge ahead.
• Parag Agrawal, Elon Musk’s predecessor at Twitter, said he may sue the company for his legal expenses.
• The FBI said using public phone chargers could risk private data in “juice jacking” attacks.
• A report on the FTX crypto exchange said hubris and greed played a major role in its collapse last year.
The European Consumer Organisation called for an investigation into “all major chatbots” last week, after Italy became the first country to ban OpenAI’s ChatGPT because of privacy concerns. Ireland’s Data Protection Commissioner may follow suit, as could those in Germany and France. The result could be a pause in the development of large artificial intelligence models in Europe, while the US and China surge ahead.
So what? Markets all over the world are being pushed to embrace generative models despite the dangers. Europe opting to pause over safety concerns could stall innovation in a sector where it’s already lagging behind.
But such a pause is exactly what Elon Musk, Steve Wozniak (a co-founder of Apple), and thousands of other technologists have called for in an open letter coordinated by the Future of Life Institute. The Guardian reported that some of the signatories may be wrongfully included and others have been retracted, but the letter’s main arguments have still caused an uproar.
The letter asks for
- a six-month pause in the training of systems “more powerful than GPT-4”;
- governments to step in to enforce a moratorium where companies won’t act; and
- developers to work with policymakers to create a generative AI governance system.
EU countries look most likely to take up the call. The European Commission’s AI Act aims to establish, in law, that some of the risks created by companies using the technology are unacceptable, and its AI Liability Directive will adapt civil liability law to make it easier to blame artificial intelligence systems for harming people.
No equivalent legislation exists in the US, China or elsewhere.
Meanwhile in China, the internet giant Alibaba announced this week it would release a competitor to ChatGPT called Tongyi Qianwen, and has shown no signs of slowing down.
Subject to controls by the Chinese government, Alibaba plans to integrate its model across its huge e-commerce platform and its CEO, Daniel Zhang, says the industry is at a “technological watershed moment”.
The rhetoric suggests rivals should act now or risk being left behind forever.
“A race is on to get these systems out in the world and thereby gain first mover advantage,” says David Leslie, Director of Ethics and Responsible Innovation Research at the Alan Turing Institute. “It means all sorts of people are rushing to deploy, without considering the risk profile. There are real, concrete, everyday social issues being raised.”
Problems include false and misleading information, representation of racial or gender bias, and the potential for emotional manipulation (see the case of Pierre and Eliza, the chatbot).
By the numbers
20,166 and counting – signatories calling for a pause.
25,000,000 – people using ChatGPT every day.
$200 million – projected annual revenue from ChatGPT Plus.
50 – percentage increase in median valuation of generative model startups since last year.
30 – factor by which Alibaba increased funding for its model training this year.
To pause or not to pause? Pausing the development of foundational models would delay the benefits they could bring. Equally it could buy time to develop legal guardrails around the technology to protect users from harm.
As the industry ponders, “the general population is being used as guinea pigs,” says Dr. Carissa Veliz of Oxford’s Institute for Ethics in AI. “OpenAI is releasing the technology into the world and letting society deal with it, with very little supervision.”
Jessica Cecil of the Reuters Institute for the Study of Journalism says what’s needed is “big multi-stakeholder coalitions to establish a value system that can be built into the rules governing what is most immediately harmful with these technologies”.
Even Sam Altman, CEO and co-founder of ChatGPT-creator OpenAI, has said such a system is a necessary ingredient for creating a “good” future with increasingly sophisticated artificial intelligence.
Moratoria on dangerous and unpredictable technologies have happened in the past. The UN gained widespread support for a pause on human cloning, for example. The debate now is about cloning human minds rather than bodies, but the question remains: how obvious must the danger become before artificial intelligence is viewed in a similar light?
To Mumbai
Apple is opening its first retail location in India – a symbol of the company’s pivot away from China. The new store will be located in Mumbai’s main business district, in a shopping complex owned by Mukesh Ambani, an Indian billionaire. The news follows an announcement last month that more iPhones will be manufactured in India’s Karnataka region after a $700 million investment in a new site there, and controversy surrounding the production of phones in China under its zero-Covid rules. Apple is busy diversifying its production processes and reducing its reliance on China. India is set to benefit – as a manufacturer but also as a market for the world’s most valuable company.
Please don’t stop the music
An authority in Italy is probing Meta about anticompetitive behaviour. The watchdog is investigating whether Meta abused its market dominance in talks over the rights to music posted on Meta’s platforms. Meta has halted talks for the renewal of expired contracts by eliminating from its social platforms all music content protected by SIAE – the Italian society of authors and publishers. According to the regulators, Meta didn’t provide SIAE with the information needed for “fair and transparent” negotiations. As a result Meta has failed to reach a copyright renewal deal with SIAE and all songs on SIAE’s books are blocked on Meta’s platforms, including Facebook, Whatsapp and Instagram.
To Japan
Microsoft invested $13 billion in OpenAI because it believed foundational models like Chat GPT would supercharge demand for other parts of its business, including its Bing search engine, GitHub coding platform, Microsoft Office suite and, especially, its Azure cloud computing business. Roughly 10 million requests are made to ChatGPT every day and that number is set to grow as OpenAI’s CEO Sam Altman courts the Japanese Prime Minister Fumio Kishida this week. His plan is to create services for the Japanese government and expand adoption among Japanese users. Reuters says Altman and Kishida discussed the risks and potential rewards of greater use of foundational models. Japan faces a crisis of labour shortages and low productivity which could make generative artificial intelligence an attractive tool for Kishida’s government.
Douyin up
Another triumph for Tencent’s main rival, ByteDance: the TikTok-creator posted a 79 per cent increase in profits this quarter, meaning it outperformed Tencent and other internet companies including Meta and Amazon. Growth in digital advertising revenues, which was reported by Bloomberg last week, fuelled the profit figure, and it was Douyin – the Chinese short-form video sharing app – rather than TikTok which proved to be the cash cow for ByteDance. Douyin operates exclusively within China, and is not subject to any of the criticisms or suspicions swirling around TikTok in the US.
Bard updates
Google has added a new feature to Bard – its “experimental conversational AI service” – which shows users how changes are being made to the model and notifies them of bugs that have emerged and been fixed. Google has stressed since Bard’s launch that public feedback will help to improve it, and that mistakes will happen in the meantime. The “experiment updates” are a step towards cautioning users about those mistakes and may encourage Bard users to be more wary of trusting its output. Whether the public should be used as testers in this way for new and potentially harmful technologies remains open to question.
Where’s Amazon’s GPT?
Amazon is a leading artificial intelligence company with a long track record in machine learning and predictive analytics. But its ChatGPT competitor is conspicuously absent from the market – an absence that says much about how OpenAI has shaped expectations of what the artificial intelligence market should look like. AWS is one of the largest artificial intelligence infrastructure and service providers on the planet, but has yet to talk about its work on natural language processing or other transformer-based models. It doesn’t have a Bard, a Bing, or a ChatGPT. But it seems an announcement is coming. The Washington Post says Amazon has told staff that it has “a lot happening in the space” and will soon share plans for its foray into the generative artificial intelligence future.
Thanks for reading. Please tell your friends to sign up, send us ideas and let us know what you think. Email sensemaker@tortoisemedia.com.
Luke Gbedemah
Serena Cesareo
