ChatGPT creator and CEO of OpenAI Sam Altman testifies in US Senate - BBC News

Marianna Brady

Live reporter, Washington DC

The US Senate committee hearing about the impacts of artificial intelligence has come to a close, and we’ll be ending our live coverage.

Today's reporting was brought to you by our technology and live teams in San Francisco, Washington DC, and the UK.

Click here to read more of the BBC's coverage on AI.

Zoe Kleinman

Technology editor

It was as much about the lawmakers trying hard not to let AI become a runaway train as it was about the unflappable Sam Altman answering question after question, many of them similar in theme.

For now at least they appear to be on the same team.

Altman said he doesn't make money from OpenAI and he set it up to try to protect humanity from less scrupulous companies. Every time someone uses ChatGPT it costs the firm money.

He does worry about the tech's potential to cause harm and he says its safety measures are in part designed to protect children even though the project is not intended for them.

The committee said it intended to make this a series of hearings and given that this one comfortably lasted three hours it's clear there is plenty more to discuss.

US congressional hearings can be combative and raucous at their worst, with lawmakers often performing for the cameras in the room or talking over witnesses.

That was not the case at today's Senate committee hearing.

Democrats and Republicans got along and even complimented each other on the collegial atmosphere in the room.

Apart from a few pointed questions, they were also largely warm and accommodating toward the three witnesses in the room - all agreeing that regulation on AI was needed.

More hearings are to follow, and it's unclear if the tone will remain as warm next time.

Speaking to reporters after the hearing, Senator Blumenthal says Sam Altman "seems to be pretty sincere".

He stressed that "Congress cannot be the gatekeeper" of regulating AI.

When asked about the regulatory agency that was discussed several times throughout the hearing, the chair said it needs to be broader than just AI.

"The Federal Trade Commission (FTC) doesn't have the capability right now" to regulate AI, he says. So someone else needs to do it, he concedes.

Tom Gerken

Technology reporter

Outside of ChatGPT, OpenAI is also responsible for incredibly sharable technology that creates images from words.

Dall-E 2 is an AI image generator. The name combines painter Salvador Dali and Pixar character Wall-E.

When Microsoft acquired a large stake in OpenAI, it began using the tool as the basis for Bing’s own image creation platform.

Dall-E 2 is just one of many image creation tools out there, including rivals Stable Diffusion and Midjourney - the latter of which boasting the most popular Discord server in the world with 15 million members.

But some have criticised the rise of AI art, fearing it will put digital artists out of business and criticising the fact that these models only function by first analysing digital art which currently exists - raising ethical questions around copyright.

Others have praised it, saying that AI art is actually creating jobs by allowing people to create things they previously would not have been capable of making.

Sam Altman told the committee today that OpenAI is working on a copyright system to compensate artists whose art was used to create something new.

"Creators deserve control," he said.

Altman also said he believed regulation should demand images state when they have been generated by AI.

But several senators asked if the company was doing enough, and one questioned why OpenAI couldn't implement such a thing immediately.

Find out more from BBC Click on the future of AI Art by clicking here.

The manner in which artificial intelligence is being used and how it can be regulated were the key focus of today's hearing.

Here are some key moments we heard:

• Richard Blumenthal, chairman of the Senate subcommittee on privacy, played a statement manipulated to sound like him by AI as he opened on the potential harms of the technology
• OpenAI CEO Sam Altman said that AI has the potential to solve "humanity's biggest challenges, like climate change and curing cancer"
• Altman and his fellow witnesses agreed that government regulation is "crucial" to the future of the industry
• AI-skeptic professor Gary Marcus said the US and others had "acted too slowly with social media" regulation but has choices to make right now regarding AI
• Marcus proposed a new cabinet-level agency to regulate the sector, an idea that Altman also appeared to back
• Altman said AI advancements will impact the workforce significantly but current models remain "a tool, not a creature"
• Asked about AI's impact on music, Altman said that content creators should have a say in how their voices, likenesses or copyrighted content are used to train AI models
• Christina Montgomery, IBM's chief privacy and trust officer, said the EU's rules on AI are "regulating by context" and provide a good lead for the US to follow
• Blumenthal said AI companies should lead with a "do no harm" approach but conceded there would be no pause in AI development until regulators can catch up

The Senate hearing has just ended.

Richard Blumenthal, the subcommittee chair, thanks the three witnesses again for their testimony.

"You've performed a great service by being here today."

He says the hearing has raised several questions about AI to promote further discussion.

The hearing appears to be wrapping up, but before it does we have heard again from Senator Hawley.

He brings up an open letter, signed by major tech players including billionaire Elon Musk.

Back in March, several AI leaders used the letter to call for a pause on AI developments more advanced than the current version of ChatGPT.

Hawley asks each of the witnesses if there should be a general pause on developments for six months.

Their answers are non-committal.

Senator Blumenthal then weighed in and said a moratorium on new developments would be a problem.

“The world won’t wait,” he says.

“Sticking our head in the sand” is not the answer.

US SenateCopyright: US Senate

Senators have finished their questions to the witnesses, and the subcommittee chair Richard Blumenthal begins to bring the hearing towards a close, during which he discusses some of the issues that he says weren’t raised.

These include monopolization, and what he calls the “dominance of markets” that excludes new competition and therefore hinders innovation and invention.

An issue which was not touched upon with “huge implications”, he says, is that of national security. He says the threats posed by America's adversaries are “very real and urgent” and need to be dealt with.

He also touches on some of the challenges around the suggestion to form a new regulatory agency, put forward by Altman.

Senator Blumenthal says that “you can create 10 new agencies”, but if you don’t provide them with the resources – not just financial, but scientific expertise as well – then private companies will “run circles around” the government.

He says “there are some really hard decisions” that need to be made around this issue, but that a “do no harm” approach must be taken, to ensure rules are introduced that are both effective and enforceable.

Shiona McCallum

Senior technology reporter

AI is much more than chatbots and many industries are embracing its capabilities.

Despite initial fears over cheating, in education we’ve seen it assist students with dissertations and give them inspiration for projects. We have also seen it have the ability to adapt to individuals needs and become their personal AI tutor.

In healthcare, AI has the potential to improve diagnostics by predicting diseases. I’ve heard some doctors are even using ChatGPT to help empathically explain medical scenarios to patients and their loved ones.

In agriculture, AI is being deployed to analyse weather and soil conditions. It can give advice on best harvesting times and recommend the right pesticides to prevent crop disease.

We’ve seen AI helping disabled people too. The models are being trained to help with speech recognition and some have developed audio descriptions for videos, making them more accessible to people who are blind and visually impaired.

James Clayton

North America technology reporter

Altman has just described in detail how he thinks this technology should be regulated.

He wants an agency to have the power to give licenses to AI companies and also take them away.

That would give a regulatory body real teeth.

Altman says the regulator should be responsible for outlining a series of safety rules that the industry has to abide by.

He also wants independent audits on AI tech companies - done by people outside of tech and the government.

This sounds magnanimous and open - but it’s worth at least flagging there can be ulterior motives when a company actively asks for regulation.

This can happen when a company thinks regulation is inevitable, and wants to get ahead of the curve by proposing legislation.

Counterintuitively, it’s also not a given that Big Tech is de facto against regulation.

The more rules and laws that apply to an area of tech the greater the barrier of entry to other start-ups. Regulation done wrong can further entrench monopolies.

Zoe Kleinman

Technology editor

Okay, things are getting feistier.

The Senators are snappy and demanding answers they can understand.

They are quick to interrupt if a response isn’t to their liking. You get the sense people in the room are getting a little bit restless.

We’ve moved from the lawmakers questioning themselves as to whether they are up to the huge task of trying to regulate AI at a fast enough pace, to those same lawmakers challenging the industry experts to lay out exactly what infrastructure they think should be put in place, and then criticising their ideas.

There’s been a lot of discussion around whether there should be a US agency, or a global body, or both.

Issues like copyright and the right to sue if an AI product causes harm have also been raised.

The senators are trying to leave no stones unturned but we all know that new tech has a habit of showing up new stones you hadn’t considered - and those can be the ones which cause the most damage.

Republican Senator Kennedy asks the witnesses to explain “in plain English” three or four reforms they would introduce if they were “queen or king for the day”.

Montgomery, from IBM, doesn’t get far into her response – talking about the importance of transparency – before Kennedy shoots back asking for clarification on what she means.

She talks about the importance of tech transparency, organisational transparency, and clear rules and clarifications.

Senator Kennedy then turns to Professor Marcus.

“This is your shot, man,” he says. “Talk in plain English – I’m looking for specificity.”

Professor Marcus complies, saying he would introduce a “nimble monitoring agency” that can monitor if the AI is being used responsibly. He also speaks about he importance of AI research.

Senator Kennedy appears satisfied with his response, and turns his attention to Altman.

Altman says he would introduce an agency that ensures compliance.

Senator Kennedy, who is well known on Capitol Hill for his colourful questioning, asks Sam Altman if he makes a lot of money in his role as CEO.

Altman smiles and notes he is "paid enough for health insurance" but has "no equity in OpenAI".

"You need a lawyer or an agent," Kennedy jokes.

"I'm doing this because I love it," replies Altman.

Altman has been an investor in the tech industry for decades, and though we don't know his exact net worth, it's fair to say he isn't struggling for cash.

He personally donated $250,000 to Joe Biden during the 2020 election.

There has been a lot of discussion about regulation in the room, but exactly what to regulate, and how it would work, are the key questions that remain.

Senators have suggested particular topics around privacy, questioning the data that AI models are trained on.

Gary Marcus, a professor with expertise in AI, told the committee he believed the best way to regulate would be to create a new government agency altogether.

“My view is we probably need a cabinet-level organisation within the US to address this,” he said.

But Christina Montgomery, IBM’s Chief Privacy and Trust Officer, disagreed.

“We don’t want to slow down regulation to address real risks right now,” she said. “We have existing regulatory authorities in place who have been clear that they have the ability to regulate in their respective domains.”

Democratic Senator Chris Coons responded that these organisations were not funded well enough to do this.

And Sam Altman, CEO of OpenAI, has taken the perhaps unusual stance of agreeing that regulation is required, and called on US Senators to regulate artificial intelligence.

“We think that regulatory intervention by governments will be critical to mitigating the risks of increasingly powerful [AI] models,” he said.

Democratic Senator Chris Coons says that he is worried about the potential use of AI to undermine faith in democratic values and free society.

He asks Montgomery if the US should follow the EU's lead.

Montgomery replies that the Europe AI Act amounts to "precision regulation" or "regulating by context" - which we'll expand upon later, as they are quite confusing terms.

There are "different rules for different risks", making it a good model for the US to base future regulatory action on, she says.

Zoe Kleinman

Technology editor

Witness Christina Montgomery is speaking about Europe's AI Act - but what is it?

The act, which is due to be voted on by lawmakers soon, is strict.

It proposes a complete ban on facial recognition tech in public places, and varying levels of rules depending on the impact of the tool in question – so for example an email spam filter would face less scrutiny than a tool for diagnosing a medical condition.

The US, on the other hand, has so far opted for guidelines and recommendations rather than bans.

The UK is trying to position itself somewhere in between. In the early days of social media, tech firms insisted they could regulate themselves and didn’t need government intervention.

We all know how that turned out. But that lesson has been learned and I think we will now see both lawmakers and tech firms wanting to avoid the mistakes of the past and their unintended, but toxic, consequences.

OpenAI – backed, lets not forget, by billions of Microsoft dollars – will be lobbying hard to try to influence the incoming rules in a way which benefits the industry.

Ben Derico

Reporting from San Francisco

Following the questions from Senator Marsha Blackburn, who represents creative hub Nashville, Tennessee - it should be clear that the effects of AI are reaching into the creative community.

Text, image, even audio and video, are being generated as we speak with simple web tools through simple text prompts.

Want to see a purple elephant playing a trombone in the style of Van Gogh? Image generators like DALL-E and Stable Diffusion can do it in seconds.

But to build that image, the algorithms first need to be "trained" on millions, sometimes billions, of illustrations "scraped" from the web.

This often means the images were taken without consent, credit, or compensation to the artists who made them.

The creative community is already raising the flag, saying this violates their copyright and poses a threat to their way of life.

The presence of AI has just introduced a light moment of US state competition to the otherwise formal proceedings.

Amy Klobuchar – the Democratic senator for Minnesota – follows remarks from Marsha Blackburn, the Republican senator for Tennessee which is home to Nashville, the country music hub.

Blackburn used her time to talk about the importance of protecting music artists from AI.

Klobuchar says that while Blackburn was speaking she used ChatGPT to search for the top creative song artists of all time.

“Two of the top three were from Minnesota,” she says, referring to Prince and Bob Dylan.

She then moves on with her questioning. "On a more serious note..."

We are now hearing from other members of the Senate subcommittee.

Marsha Blackburn, a Republican from Tennessee, asks how musicians and other content creators can expect their works to be used to train future AI models.

She notes that OpenAI's Jukebox - which generates music in raw audio form when you input an artist, genre or lyrics - has the capability to create "re-renditions" of songs by the country music artist Garth Brooks.

Sam Altman replies that content owners should be able to receive "significant upside benefit" if their works are used to train AI models.

He adds they should also have the power to prevent the use of their copyrighted songs and images, as well as their voices and likenesses.