Claim: The U.S government acquisition of Intel shares should be treated as a weak indicator of how important it sees the future strategic importance of AI.
It is (usually) obvious to determine how the government feels when the issue is directly political by looking at the beliefs of the party in charge. This is a function of how the executive branch works. When appointing the head of a department, the president will select someone who generally believes what they believe, and that person will execute actions based on those beliefs. The “opinion” of the government and the opinion of the president will end up being essentially the same in this case. It is much harder to determine what the government as a whole’s position is when the matter is not directly political. Despite being an entity comprised of hundreds of thousands of people, the U.S as an entity certainly has weak/strong opinions on almost all issues. Think rules and regulations for somewhat benign things, or the choices and tradeoffs made during a disaster scenario. Determining this opinion can be very important if something you are doing hinges on the way the government will act in a scenario, but can be somewhat of a dark art without historical examples to fall back on or current data on what actions they have taken so far. If we want to determine the government’s position on AI, the best thing we can do is try to look for indicators via their direct actions relating to AI.
The government acquisition of 10 percent of Intel, to me, seems like an indicator of the government’s opinion on the importance of AI. The stated reason for the acquisition was, paraphrased, “We gave Intel free money with the CHIPS act, and we feel that doing so is wrong, so we decided to instead give all that awarded money + a little more in exchange for some equity so America and Americans can make money off it”. I don’t think this is wholly untrue, but it feels incomplete and flawed to me. The government directly holding equity in a company is a deeply un-right-wing thing to do, and the excuse of “the deficit” feels weak and underwhelming to completely justify such a drastic action. I find it plausible that certain people in the government who have political power but aren’t necessarily public-facing pushed this through as a method to ensure closer government control of chip production in the event that AI becomes a severe national security risk. Other framings are possible, such as the idea that they want chip fab in America for more benign reasons than AI as a security risk, but if so then why would they need to go so far as to take a stake in the company? The difference between a stake and a funding bill like the CHIPS act is the power that stake gives you to control what goes on within the company, which would be of key importance in a short-medium timeline AGI/ASI scenario.
I believe this is a far stronger indicator than the export controls on chips to China or the CHIPS act itself. It’s simplified but probably somewhat accurate to consider the cost of a government action as the monetary cost + the political cost, with political cost being weighted more strongly. Simple export controls have almost zero monetary cost and almost zero political cost, especially when they are for a hyper-specific product like a single top-end GPU. The CHIPS act had a notable monetary cost, but almost zero political cost (most people don’t know that the act exists). This scenario has a small or negative monetary cost (when considering the CHIPS act money as a sunk cost), but a fairly notable political cost (see this Gavin Newsom tweet as evidence for this, along with general sentiment among conservatives about this news).
I acknowledge this as a weak indicator, but I believe looking for any indicators of the governments position on the issue of AI has value in determining the correct course of action for safety, policy especially.
Claim: The U.S government acquisition of Intel shares should be treated as a weak indicator of how important it sees the future strategic importance of AI.
It is (usually) obvious to determine how the government feels when the issue is directly political by looking at the beliefs of the party in charge. This is a function of how the executive branch works. When appointing the head of a department, the president will select someone who generally believes what they believe, and that person will execute actions based on those beliefs. The “opinion” of the government and the opinion of the president will end up being essentially the same in this case. It is much harder to determine what the government as a whole’s position is when the matter is not directly political. Despite being an entity comprised of hundreds of thousands of people, the U.S as an entity certainly has weak/strong opinions on almost all issues. Think rules and regulations for somewhat benign things, or the choices and tradeoffs made during a disaster scenario. Determining this opinion can be very important if something you are doing hinges on the way the government will act in a scenario, but can be somewhat of a dark art without historical examples to fall back on or current data on what actions they have taken so far. If we want to determine the government’s position on AI, the best thing we can do is try to look for indicators via their direct actions relating to AI.
The government acquisition of 10 percent of Intel, to me, seems like an indicator of the government’s opinion on the importance of AI. The stated reason for the acquisition was, paraphrased, “We gave Intel free money with the CHIPS act, and we feel that doing so is wrong, so we decided to instead give all that awarded money + a little more in exchange for some equity so America and Americans can make money off it”. I don’t think this is wholly untrue, but it feels incomplete and flawed to me. The government directly holding equity in a company is a deeply un-right-wing thing to do, and the excuse of “the deficit” feels weak and underwhelming to completely justify such a drastic action. I find it plausible that certain people in the government who have political power but aren’t necessarily public-facing pushed this through as a method to ensure closer government control of chip production in the event that AI becomes a severe national security risk. Other framings are possible, such as the idea that they want chip fab in America for more benign reasons than AI as a security risk, but if so then why would they need to go so far as to take a stake in the company? The difference between a stake and a funding bill like the CHIPS act is the power that stake gives you to control what goes on within the company, which would be of key importance in a short-medium timeline AGI/ASI scenario.
I believe this is a far stronger indicator than the export controls on chips to China or the CHIPS act itself. It’s simplified but probably somewhat accurate to consider the cost of a government action as the monetary cost + the political cost, with political cost being weighted more strongly. Simple export controls have almost zero monetary cost and almost zero political cost, especially when they are for a hyper-specific product like a single top-end GPU. The CHIPS act had a notable monetary cost, but almost zero political cost (most people don’t know that the act exists). This scenario has a small or negative monetary cost (when considering the CHIPS act money as a sunk cost), but a fairly notable political cost (see this Gavin Newsom tweet as evidence for this, along with general sentiment among conservatives about this news).
I acknowledge this as a weak indicator, but I believe looking for any indicators of the governments position on the issue of AI has value in determining the correct course of action for safety, policy especially.
cuda support for Intel GPUs sucks, is Trump not aware?