Scarlett Johansson's AI controversy is reminiscent of the bad old days of Silicon Valley

The motto “act fast and break things” continues to haunt the tech industry nearly two decades after it was coined by a young Mark Zuckerberg.

These five words have come to symbolize Silicon Valley at its worst: a combination of ruthless ambition and breathtaking arrogance, profit-driven innovation without fear of consequences.

I was reminded of this phrase this week when actress Scarlett Johansson clashed with OpenAI. Johansson claimed that both she and her agent declined to let her voice ChatGPT's new product, but then when the product was announced it sounded exactly like her voice anyway. OpenAI denies that it was a deliberate copycat.

This is a prime example of why the creative industries are so concerned about being copied and eventually replaced by artificial intelligence.

Sony Music, the world's largest music publisher, wrote to Google, Microsoft and OpenAI last week demanding to know whether its artists' music has been used to develop AI systems, saying permission had not been granted.

There are echoes of the macho corporate Silicon Valley of old: the unofficial business plan: asking for forgiveness, not permission.

But tech companies in 2024 will be very keen to distance themselves from that reputation.

OpenAI is not built in that mold. It was originally established as a non-profit organization that invested the excess profits invested back into the business.

When it launched its for-profit arm in 2019, the company said it would be led by the nonprofit arm and that there would be caps on the returns investors could earn.

Not everyone was happy with the change. That was said to be the main reason original co-founder Elon Musk decided to withdraw.

When OpenAI CEO Sam Altman was abruptly removed from his board of directors late last year, one theory was that he wanted to move further away from his original mission. I didn't know for sure.

But even as OpenAI becomes more profit-driven, it still must face its responsibilities.

Nearly everyone in policymaking circles agrees that clear boundaries are needed to get companies like OpenAI to cooperate before disaster strikes.

So far, the AI ​​giants have acted largely on paper. Six months ago, at the world's first AI Safety Summit, many technology leaders made a voluntary pledge to create responsible and safe products that maximize the benefits and minimize the risks of AI technology. I signed it.

These risks, initially identified by event organizers, were a nightmare. At the time, I asked about the more real threat of AI tools discriminating against people or forcing people out of their jobs, and I was told that this gathering was only meant to discuss worst-case scenarios, and that the Terminator, Doomsday, etc. We were told flatly that this was a realm in which AI would go out of control and destroy humanity.

When the summit resumed six months later, the word “safety” had been removed from the conference title entirely.

Last week, a draft UK government report from a group of 30 independent experts concluded that there was “no evidence yet” that AI could create biological weapons or carry out sophisticated cyber attacks. The possibility of humans losing control of AI is “highly debatable,” the magazine said.

Some people in the field have been saying for quite some time that the immediate threat from AI tools is that they either take away jobs or become incapable of recognizing skin color. Dr. Raman Chaudhry, an expert on AI ethics, says these are “real questions.”

The AI ​​Safety Institute declined to say whether it has performed safety testing of new AI products that have been launched in recent days, specifically OpenAI's GPT-4o and Google's Project Astra, both of which are among the most powerful and advanced generative AI systems available to the public that I have seen to date. Meanwhile, Microsoft announced a new laptop with AI hardware, marking the beginning of AI tools being physically built into devices.

The independent report also found that there is currently no reliable way, even among developers, to understand exactly why an AI tool produces the output it does, and that evaluators intentionally induce AI tools to malfunction. It also notes that there are no best practice guidelines for the established safety testing method known as red team testing.

At a follow-up summit co-hosted by Britain and South Korea in Seoul this week, companies pledged to shelve products that do not meet certain safety standards, but these will not be set until the next meeting. 2025.

Some worry that all these promises and commitments are not being fully delivered on.

“Voluntary agreements are essentially just a way for companies to score their own agendas,” said Andrew Straight, deputy director of the independent Ada Lovelace Institute. “This is not a substitute for the legally binding and enforceable rules needed to encourage the responsible development of these technologies.”

OpenAI just announced a unique 10-point safety process that the company says it is working on, but one of its senior safety-focused engineers recently resigned and told X that his department was “working against headwinds.” “I'm sailing along,” he wrote.

“Over the past few years, safety culture and processes have taken a backseat to shiny products,” posted Jan Leike.

Of course, there are other teams at OpenAI who continue to focus on safety and security.

However, there is currently no official independent oversight of what they are actually doing.

“There's no guarantee these companies are keeping their promises,” says Professor Wendy Hall, one of Britain's leading computer scientists.

“How can we hold them accountable for what they say, just like we do with pharmaceutical companies and other high-risk sectors?”

We may also find that these powerful tech leaders become less compliant once pressure mounts and voluntary agreements become a bit more legally enforceable.

When the British government said it wanted powers to allow big tech companies to suspend the rollout of security features if they thought they might compromise national security, Apple called it an “unprecedented overreach” by lawmakers and threatened to remove its services from the UK.

The bill passed, and for now, Apple remains in business.