Generative AI risks concentrating the power of Big Tech. Here’s how to stop it.

This story originally appeared in The Algorithm, our weekly AI newsletter. To get stories like this delivered to your inbox first, sign up here.

If regulators don’t act now, the generative AI boom will further concentrate the power of Big Tech. That’s the central argument of a new report from the AI ​​Now research institute. And it makes sense. To understand why, consider that the current AI boom depends on two things: large amounts of data and enough computing power to process it.

These two resources are really only available to large companies. And while some of the more exciting apps, such as OpenAI’s ChatGPT chatbot and Stability.AI’s Stable Diffusion AI image generation, are created by startups, they rely on agreements with Big Tech that allow them to provide access to its vast data and computing resources.

“A few big tech companies are on the verge of consolidating their power with AI rather than democratizing it,” says Sarah Myers West, chief executive of the AI ​​Now Institute, a nonprofit research organization.

Right now, Big Tech has a grip on AI. But Myers West thinks we are actually at a watershed moment. It’s the start of a new hype cycle, which means lawmakers and regulators have a unique opportunity to make the next decade of AI technology more democratic and fair.

What sets this tech boom apart from previous ones is that we have a better understanding of all the catastrophic ways AI can go wrong. And regulators around the world are paying close attention to it.

China just unveiled a generative AI bill calling for more transparency and oversight, while the European Union negotiates the AI ​​law, which will force tech companies to be more transparent about how AIs work. generative AI systems. It also provides for a bill to make them liable for damage caused by AI.

The United States has always been reluctant to regulate its technology sector. But that is changing. The Biden administration is seeking information on ways to oversee AI models such as ChatGPT, such as requiring tech companies to produce audits and impact assessments, or requiring AI systems meet certain standards before publication. This is one of the most concrete steps the administration has taken to limit the harms of AI.

Meanwhile, Federal Trade Commission Chairman Lina Khan also highlighted Big Tech’s advantage in data and computing power and pledged to ensure competition in the AI ​​industry. . The agency dangled the threat of antitrust investigations and crackdowns on deceptive marketing practices.

This new focus on the AI ​​industry is partly influenced by the fact that many members of the AI ​​Now Institute, including Myers West, have spent time at the FTC.

Myers West says his stint taught him that AI regulation doesn’t have to start from a blank slate. Instead of waiting for AI-specific regulations such as the European AI Act, which will take years to put in place, regulators should strengthen enforcement of existing data protection and competition laws.

Because AI as we know it today is largely dependent on massive amounts of dataData politics is also AI politics, says Myers West.

Case in point: ChatGPT has come under scrutiny from European and Canadian data protection authorities, and has been blocked in Italy for allegedly illegally mining personal data from the web and misusing personal data.

The call for regulation does not just come from government officials. Something interesting happened. After decades of fighting tooth and nail against regulation, most tech companies, including OpenAI, now say they like it.

The big question that everyone is still wrestling with is how AI should be regulated. Although tech companies claim to support regulation, they still pursue a “publish first, ask later” approach when it comes to launching AI-powered products. They rush to release AI models that generate images and text as products, even though those models have major flaws: they invent nonsense, perpetuate harmful biases, infringe copyright and contain loopholes of security.

The White House’s proposal to tackle AI liability with post-launch measures of AI products such as algorithmic audits is not enough to mitigate the damage caused by AI, according to the AI Now report. Stronger and faster action is needed to ensure companies first prove that their models are market-ready, says Myers West.

“We must be wary of approaches that do not place the burden on businesses. There are many regulatory approaches that essentially put the onus on the general public and regulators to eradicate the harms of AI,” she says.

Above all, according to Myers West, regulators must act quickly.

“There must be consequences when (tech companies) break the law.”

Deeper learning

How AI is helping historians better understand our past

It’s cool. Historians have started using machine learning to examine historic documents stained by centuries past in moldy archives. They use these techniques to restore ancient texts and make important discoveries along the way.

Join the dots:Historians say the application of modern computing to the distant past helps to establish broader connections across the centuries than would otherwise be possible. But there is a risk that these computer programs will introduce their own distortions, slipping biases or outright falsifications into the historical record. Learn more about Moira Donovan here.

Bits and Bytes

Google overhauls search to compete with AI rivals
Threatened by Microsoft’s relative success with AI-powered Bing search, Google is building a new search engine that uses large language models and upgrading its existing search engine with AI features. He hopes the new search engine will give users a more personalized experience. (The New York Times)

Elon Musk created a new AI company to rival OpenAI
For the past few months, Musk has been trying to hire researchers to join his new AI company, X.AI. Musk was a co-founder of OpenAI, but was ousted in 2018 after a power struggle with CEO Sam Altman. Musk has accused OpenAI’s Chatbot ChatGPT of being politically biased and says he wants to create “truth-seeking” AI models. What does that mean? Your guess is as good as mine. (The Wall Street Journal)

Stability.AI is in danger of sinking
Stability.AI, the creator of the open-source image-generating AI model Stable Diffusion, has just released a new version of the model with slightly more photorealistic results. But the company is in trouble. He quickly burns through cash and struggles to generate revenue, and the staff lose faith in the CEO. (Semafor)

Discover the worst AI program in the world
The bot on Chess.com, depicted as a Bulgarian man wearing a turtleneck with bushy eyebrows, a thick beard and a slightly receding hairline, is designed to be absolutely horrible at chess. While other AI robots are programmed to dazzle, Martin reminds us that even dumb AI systems can still surprise, delight and teach us. (Atlantic)

Tech