David Patterson a agenda to maximize AI benefits and reduce losses

It was the subject of an enthusiastic conversation that I did in February 2024 with Andy Convinski, one of my alumni and two AI-related startups, databricks and co-founders. Andy shared her mistrust that a friend’s son was out of his computer science program. This bright student believed that AI would soon make the programmer obsolete.

He is not alone: ​​According to Galp, three-fourths Americans say AI will reduce the total number of jobs within ten years. This reminded me of an equally promising student, who left computer science 20 years ago, offering that offshoring meant that almost all programming jobs would turn into low -income countries such as India. Their nervousness was incorrect: since 2000, the number of jobs for the programmer in the US and their inflation-proposed salary increased by half.

Our conversation turned into shared despair over the polarized discourse between AI “accelers” and “Doomers”. Reality, we agreed, more fine. We concluded that there is an urgent need for computer scientists to play a more active role in shaping steering research and story. Instead of only anticipating that the effect of AI would be given a laisez-faire approach, our goal was to propose what effect the effect could be given to maximize the effect and reduce downside.

We then assembled nine of the world’s major computer scientists and enhanced AI stars from Academia, Startups and Big Tech, to detect the impact of AI’s practical proximity. We also interviewed two dozen other experts about the impact of AI on our specialties, including John Jumper, the winner of the Nobel Prize in this year’s Chemistry, on science; President Barack Obama on rule; His former United Nations Ambassador to Security and National Security Advisor Susan Rice; And Eric Schmidt, a philanthropist and former Chief Executive Officer of Google, on many subjects. For those interested, we have compiled our learning into a more detailed 30-page paper, titled “Impact on billions of lives of Shapeing AI”.

Five guidelines emerged to exploit AI for public good. We believe that they should guide our efforts in both discovery and deployment of this transformative technology.

First of all, humans and AI systems working as a team either do more than themselves. Applications of AIs focused on human productivity produce more positive benefits than people focused on human replacement. Tools that make people more productive enhance their employment, satisfaction and opportunity. People can act as safety measures if AI closes the course in areas for which it is not well trained. In short, focusing on human productivity helps both people and AI succeed.

Second, to increase employment, the objective for improving productivity in areas that will create more jobs. Despite the tremendous productivity benefits in computing and passenger aviation, the US had 11 times more programmer and eight times more commercial-airyline pilots in 2020 than in 1970. This growth is because there are programming and air transport sectors, for which economists say, the demand is elastic. , On the other hand, agriculture is relatively disqualified, so productivity gain meant that the number of agricultural jobs fell three-fourths in a human lifetime (1940 to 2020). If AI physicians aim to improve productivity in elastic areas, despite the fear of the public, AI can actually increase employment.

Third, the AI ​​system should initially aim to remove intoxication from existing tasks. Releasing time for more valuable work will encourage people to use the new AI Tool. Doctors and nurses choose their career as they want to help patients, not endless documentation. School students prefer not to teach, grading and record-keeping. AI tools should be given high priority which are going to improve the significance of the current work of people in hospitals and classes.

Fourth, the effect of AI varies by geography. Eric Schmidt emphasized that rich countries are concerned about AI that they are displacing high trained professionals, but countries with lean economies face shortage of skilled experts. AI can provide such expertise more widely in such areas, potentially enhances the quality of life and economic development, as mobile phones have become. For example, an AI system that improves skills and productivity of nurses and physician assistants will provide more patients in high quality health care that are lack of doctors. The growing popularity of smartphones in lower and medium-or-countries enables widespread access to multilingual AI models that can help people dramatically reach information, education, media/entertainment in low and medium-income countries, And if more desired in their natives, then languages. Improvement in local economies and important services can also provide options for migration for some in moderate income countries.

And finally, we need better matrix and methods to evaluate AI innovations. Sometimes a marketplace can do this, such as for a professional programmer for AI Tools. This cannot do this in high-domain domains, as we cannot take the risk of damaging the participants. We need to use gold-standard devices: A/B tests, random controlled tests and natural experiments. It is equally essential to evaluate post-duploological monitoring as to what AI innovation they say that they say what they are doing, whether they are safe, and they have outdoor. We also need to continuously measure the AI ​​system in the area so that they can be able to grow.

There is no dearth of concerns about the risks and complications of AI, which we address in long paper: data privacy and security, intellectual-property rights, prejudice, information accuracy, threat to humanity from more advanced AI, and Energy consumption (though (although at this last point, AI accounts under a quarter of 1% of global power use, and the International Energy Agency increases the AI’s estimated energy consumption for 2030 modestly relatively relative to other trends Has been) agrees.

Although there are risks, there are many opportunities, both are known and unknown. Ignoring the benefits of AI can be as big mistake as it is to ignore its risks. AI moves quickly, and governments should maintain momentum. In order to cooperate with the government with the government in the successful development and deployment of chips and cars, the government proposed a coordinated public-private partnership for AI. Its goal will be to remove bureaucracy obstacles, ensure security and provide transparency and education to policy makers and public.

At this point, readers can expect that we are going to ask scientists for government money. But we believe that for these efforts, money should come from philanthropy of technologists who are rich in computer industry. Many have already promised support, and we hope to join more. We feel that these commitments should be deployed in two ways: to encourage research and identify successes, and to make major motivated awards to fund three to five years of multi-disciplinary research centers.

We considered an AI Moonshot. But which goal? We can create an AI mediator that orkestrates the conversation in political chass to get us out of polarization and pull back into pluralism. We can take advantage of the growing spread of smartphones by aiming to create a tutor app for every child in our language, for our culture, and in our best learning style. We can enable biologists and neurocystists to progress a century in a decade. But if we make the right blueprint for innovation, and bring experts and users together, we do not just need to take a moon.

David Patterson is a Pardi professor of computer science, a reputed engineer at Emeritus and Google at the University of California at Berkeley.

© 2025, The Economist Newspaper Limited. All rights reserved. From The Economist, published under license. The original material can be found on www.economist.com