I write this on a plane back in Washington, DC, from a conference in the Bay region, the country of tomorrow. While the conference It was not a question of AI, it is the region of the bay, and therefore around 90% of the conversations concerned AI.
It is difficult to overestimate the extent of the gap between the cultures of the Bay and DC region on this subject. AI has certainly become a real part of the political conversation at DC, but only in a fairly technical way, in the short term and not particularly very publicized: how should we regulate deep counterfeits? How should we Manage growing requests for energy data centers? If we Demand that Nvidia processors have a small component who can know if the chip is physically in China To prevent Beijing from getting your hands on too much?
But if the concerns of DC AI are daily, the bay region is existential.
In Berkeley, or at least among the crowd I was talking about, the questions were more like: we will never be able to prevent these machines from cheating on our attempts to assess us, to make us sing When we lead their goal, to actively working to avoid being closed? (These are all real things that researchers have found that advanced AI models can do.) If we do not solve these problems, will we survive the next 10 years?
Register here to explore the big complicated problems with which the world is confronted and the most effective means of solving them. Sent twice a week.
When questions a little less apocalyptic like “how are we going to face if billions of people are suddenly unemployed due to AI and the progression of robotics”, the tone of most of the answers I obtained was something like “God, I really hope that it turns out to be the biggest problem. This means that we have all survived. ”
Temperament, I am more inclined to think of these things in a very concrete and short -term way. There is a reason why I live in Washington, DC; It is a city for gaps in a good mood. Thus, of course, all the speeches on AI made me think about the huge bill of budget reconciliation adopted by the Chamber and to be considered by the Senate.
Let me be frank: it is, in a large and small way, not a budget that takes AI seriously. Worse, if you think that this technology will have a still slightly significant influence on the world during the next decade, the only great good bill will worsen this influence.
Tips directly linked to AI
There is a section of the bill directly on AI, which is the Moratorium proposed on most attempts at the level of the state to regulate the AI For the next 10 years. Originally, it was a pure and simple prohibition, but because of the limits of what the bills of reconciliation can make on non -budgetary issues – and attempt to regulate the regulations are clearly non -budgetary – it now takes the form of a requirement according to which states refrain from the regulations If they want to get large -band money.
There are reasonable arguments that AI policy should occur in a federal level rather than state. But this is not a case where the federal government has a well supported political framework which it seeks to impose instead of the policies of the States. This is a case where the federal government wishes to remove or prevent state regulations and replace them with nothing at all.
It is not surprising that the interests of companies such as the ventilation capital Andresen Horowitz are Crawling their DC lobbying effort In the midst of this fight. AI will change our lives fairly quickly. The public is already very suspicious And will want the regulations, requests that should only increase when the economic and short -term work effects of AI become palpable. The only way for the industry to prevent this is to block in a FAIRE regime at the moment. If you think there is even a chance that these systems can cause serious problems worthy of regulation, this is a very dangerous provision. Fortunately, even completely conservative Republicans in The two houses seem to achieve this, and I hope this reaction kills the layout.
Almost as relevant to industry, the provisions reduce subsidies for clean energy development. AI training and deployment requires many data centers Lots of very expensive fleas which must be executed 24/7 to reimburse their immense initial cost. These centers need power sources 24/7 just as reliable. Ideally, this comes from clean sources such as nuclear, geothermal energy or more solar batteries. A little worse would be natural gas. Many worse would be coal.
The reconciliation bill takes a number of actions to reduce the chances that data centers are supplied by clean sources. Of course reduces generous grants The law on the reduction of inflation created to encourage clean energy, which can compensate Up to 30% of the cost a new power plant.
The nuclear industry, the specific source to which the Republicans are generally more friendly, has warned that the cuts could make them seriously Also. The bill too Take an ax of war at the loan programs officeA tool from the energy department to invest in clean energy which is particularly important for nuclear and geothermal energy. The energy secretary, Chris Wright, went to Ask the Republicans to resume nuclear and geothermal cups; I do not think that a member of the cabinet asked for smaller cuts in another section of the bill, but it was sufficiently worrying to arouse intervention.
As political analysts Thomas Hochman and Pavan Vekatakrishnan noted in the Washington Post, “Congress approach Almost specific disadvantages It works 24/7 ”, injuring them even more than wind and solar. It is almost as if it were designed to operate new data centers on dirty fuels, or perhaps to encourage companies to build them abroad.
Work requirements in a post-work world
But the big important problem with the bill is his obsession with bacon on more expensive, poorly administered and ineffective work requirements on programs like Medicaid and food coupons.
I thought it was bad policies Before AI became a big problem, and I am happy to complain about myself why. They are cruel, they Do not lead people to work moreAnd for Medicaid in particular, Even the conservatives who normally love the work requirements accept that they are completely ineffective.
But save for a second. Right away, managers of World IA companies in the world are declaring this in the decadeThey can fully automate a huge part of human work. Maybe you think they are out of their squash and that nothing like that will happen. It is possible. It is also possible that these incredibly powerful people with several billion dollars at their disposal are able to succeed what they have decided to do.
It is Also Possible that even a lot, much less powerful, like those available today, will eventually lead to a significant job loss. We see some indications It is already happening. Even in the slowest plausible chronology for AI that I can imagine, you will always have companies like Waymo to use it to move human work in specific industries.
In a world where Uber drivers and trucks are suddenly unemployed because of their own fault, adding work needs to food coupons and Medicaid is cruel. This will not make them find a job, at least in the short term; The work in their vocation has disappeared. Perhaps they should change the professions-but are we really convinced that their new job will not be automated in the same way? Do they not need help during their transition?
Vice-president JD Vance made a speech in March Where he recalled the steel plant in his hometown of Ohio, saying: “It was the vital element of the city in which I grew up. When it went from 10,000 jobs to 2,000 jobs, American workers started to be destroyed in the process. We cannot continue to do so. ”
But his party’s budgetary bill does exactly this. He sees people whose livelihoods could be destroyed with imminence and actively Remove support from them. “We can’t keep doing this”? You do that right away.
In a truly transformer AI world, automating 10 or 20 or perhaps even 100% of human work, work requirements pass from cruel to a combination of cruel, bizarre and idiots. They would be as if the congress was today to adopt a dedicated law establishing labor standards for drivers of horse and bouggy. Imagine people in a world of transformer “you have to work to get food coupons”. Work? What work? Unemployment is 30% and increasing, what are you even talking about?
David Sacks, a venture capital and one of Trump’s closest advisers on AI, has generally been disdainful about the AI potential to threaten jobs. But even conceded him on a recent episode of sound All Podcast, “if there is a generalized disruption of employment, then obviously The government will have to react and we will be in a very different societal order. »»
At the same time, on X, it is declaring“The future of AI has become a Rorschach test where everyone sees what they want. The left envisages a post-economic order in which people stop working and rather receive the advantages of the government. In other words, everyone on well-being. It is their fantasy; it will not happen.”
Very well, you don’t want that. But the AI will certainly move many jobs, if not eliminate them, and the bags itself admit that you need massive government intervention in this case. I do not have a clear idea what this intervention would ideally look like; We know so little about how this technology will distribute in society, how speed it will improve and what it means for jobs. It is an area that needs much more attention, IA companies, governments and civil society.
But I feel confident on a point. AI will make a little more precarious. The professions will be threatened. People will lose their jobs. The questions are of how many of them, and if and how speed they will get new ones.
Given all this, adding new work requirements to safety net programs is not only cruel or reckless. It is a sign that this administration, and its technological advisers like Sacks, do not take the future of AI at all seriously.
