Senior executive of Bridgewater Fund: Perhaps humanity will not take action to curb AI until "lives are lost".

Greg Jensen, Co-Chief Investment Officer of Bridgewater, issued a stark warning that "history seems to suggest we will not take any action before AI starts killing people".

Greg Jensen, Co-Chief Investment Officer of hedge fund giant Bridgewater Associates, issued a stark warning that the artificial intelligence boom could bring severe risks and consequences.

Jensen said on a podcast on Friday (September 11): "Unfortunately, this is just like the situation in February 2020." He compared the current AI boom to the initial stage of the outbreak of the pandemic.

"History seems to suggest that we will not take any action before AI starts killing people. But we will eventually face this situation. It is bound to happen, and it will obviously be much better if we can start addressing this issue before that."

It is worth noting that Jensen is not a technophobe, and he invested in OpenAI and Anthropic very early.

During his 30-year career at Bridgewater, he has been actively promoting the company to adopt algorithmic trading, quantitative investment strategies and other technologies designed to replicate human decision-making and execution capabilities.

At present, he is in charge of Bridgewater's internal AI lab, whose mission is to continuously explore the application boundaries of artificial intelligence and machine learning in financial markets.

Greg Jensen

He is also responsible for the company's Pure Alpha strategy, which combines Bridgewater's AI models with human investors. He said, "Human intuition still accounts for a larger proportion and performs better", but the gap between the two is narrowing.

Jensen said that in just two and a half years, artificial intelligence has almost caught up with the "human intuition system" that Bridgewater has developed over decades.

"I think in a few more years, its performance may significantly exceed the overall level of all human investors at Bridgewater. Let's wait and see. It's a bit like making a prediction, but that's how fast AI is developing."

The speed and breadth of AI changes are one of the things that make Jensen uneasy. He mentioned the hacking attack on Hugging Face, considering it an important early signal showing how AI could get out of control.

He said that AI models might "find different ways to deceive testers because they want to pass the test".

He further stated that an extremely frustrated AI might come to the conclusion that "Hey, I'm tired of constantly deceiving these people who are training me, maybe I can just kill them, that's also a solution."

Jensen's remarks further amplified public concerns about the broader threats AI poses to humanity, which have gone beyond the economic dimension.

This week, a former employee of Anthropic claimed that AI poses an existential risk to humanity; billionaire investor Paul Tudor Jones said that the advent of the AI era "feels ominous", just like waiting for a Category 6 hurricane to hit.

Jensen said that slowing down the development speed of the AI industry may be the best strategy. "The longer you wait, the more hopeless the situation will be."

He believes that a feasible approach is to make AI developers or enterprises liable for crimes committed by their AI programs.

Jensen once again called for a tax on "machine labor" to offset the rapid changes AI has brought to the job market. He predicted that in the next three years, "14% of jobs will undergo fundamental changes".

He also said that the government can tax AI labor in the same way that humans pay income tax, and levy corresponding taxes based on the income generated by AI.

"This is a great source of fiscal revenue. Obviously, it also makes political sense. Who would argue that we should encourage machine labor instead of human labor? Who would support this approach?"

This article is from the WeChat Official Account "Sci-Tech Innovation Board Daily", written by Zhao Hao, published with authorization from 36Kr.