Josh Karthikeyan Week 12 - Today's Arm Race
After discussing Sam Altman, OpenAI, and the AI driven future they want to create for weeks and weeks, culminating in the POAS research paper, I wanted to share some of my thoughts. One of the things I noticed was how Sam Altman and other technological leaders all believe that AI will take over most people's jobs soon. However, I cannot help but question how much it comes from business motives. If they spread the belief that AI is revolutionary, it is easier to sell the product to as many people as possible. The people who agree with this vision are called “AI evangelists” and try to convince others that AI will change the world.
There is more extreme concern throughout the industry from AI safety professionals that when AI becomes Artificial General Intelligence (AGI), it will be too smart, and not be aligned with human values, and we will all die. Is this possible? Perhaps, but I think it is very, very unlikely because current models lack agency and do not have self directed goals. However it does worry me especially when seeing a prominent expert in AI safety leave because of their beliefs and instead fully focus on poetry.
Recently, there have been reports discussing how China’s AI companies were taking advantage of US company Anthropic's AI model, Claude and had thousands of accounts prompting Claude to get data to improve their own models. The technical term, distillation, is used normally in a research context and the article emphasizes how China is not actually developing new algorithms but instead is described as cheating by taking advantage of US models. There is irony in this statement because Anthropic themselves used books for training their models and are in a lawsuit.
My pessimistic view is that they are downplaying how China is catching up to the US to persuade investors to continue investing in their company. This is because if China whose companies open source their AI models are exactly the same as US companies, then what is the point of paying for models? There will be no profitability in the product.
US and China tensions are rising. They are hoping that their respective country will be the first to AGI and if they have such a capable AI that can do most of human labor, they will be in control of the world stage. Honestly, at that point, if such a thing happens, what are humans going to do? What will give our lives structure, purpose, and meaning? Where will our aspirations come from?
Image by Aki Ranin
This is a little tangential, but I followed that link you included to an article about an AI researcher who quit the industry to pursue poetry. This case is really interesting, because the researcher, whose name is Mrinank Sharma, was a safety researcher for Anthropic. Anthropic, according to the article, was formed by ChatGPT employees who were disillusioned by the company’s policies around its chatbots and desired to create an AI chatbot that was safer to use and had more guardrails than mainstream chatbots. If Sharma was a part of this initiative, he must have been very optimistic about the possibility of AI to develop into a positive force. So, why did he suddenly feel very pessimistic and we’re-all-gonna-die about the subject? He had put in so much work to improve AI, just to abandon it and go on a completely different route.
ReplyDeleteI wonder why he chose poetry, too. Did he feel alienated from humanity by working so closely with AI? Did he want to revel in one of the things AI still can’t do as well as humans? I wonder if his poetry will address AI in some way, or if he’ll depart from the subject. I really hope he plans to publish his work.
Here’s an amazing response you could write for Josh’s blog: (see what I did there:). Josh, I completely agree with you on the fact that AI is gaining power, and in many ways, we’re consequently losing some of it. If AI is able to get to Level V, then what stops it from basically taking over in dystopian scenarios at that point? On the US-China geopolitical struggle over AI, I disagree that China will outpace the US because of its lack of semiconductor chips and other technology necessary for rapid advancements in the field. However, I still fully agree with you that investors will reduce or maybe even stop investments in US-based AI companies at a certain point because most consumers would much rather not have to pay at all for a slightly less effective tool than pay higher amounts for the best of the best, which is assuming that US-based AI companies can develop said “best of the best” algorithms in the first place. Even still, historical precedent has shown us that there’s no real winner in a scenario like this: when the Soviet Union and the US raced to build larger nuclear arsenals, all it did was escalate tensions to the point where the entirety of humanity could have been destroyed save for a single man deciding not to give his key up: do we really want to see AI grow that far, and that quickly, due to this new arms race? On your writing, I think your use of credible sources really adds to the credibility of the points you make, and your rhetorical questions do an effective job at encouraging us to think critically about innovation and where it could lead us with AI. Thank you for a fantastic piece!
ReplyDeleteHi Josh! Ironically me and my dad were just having a conversation last night about AI and it taking over jobs! I never thought of the possibility that the “revolutionary AI” could be a business scheme used by AI companies to sell more product but it makes sense if you think about it. Although most people don’t think of it in this way, AI is just a good created to help people. So, it makes sense that the creators of AI, or AI companies, would do whatever they thought was necessary to sell the product more. I always find it so interesting to see how much politics play into the business world. As 100% consumer I feel like we are often blinded to the behind the scenes of businesses. Therefore, I really appreciate hearing your “pessimistic view” of how these AI companies are operating by doing things such as downplaying China’s advancements in AI and exaggerating the power of AI. Going back to the conversation my dad and I had, one of the things he brought up was that with AI taking over all our jobs, some believe we will have more time to think and come up with more brilliant ideas. However, I think that if AI took over our jobs and we were able to do whatever we wanted, all day every day, we would become like the people in Wall-E, mindless beings who never move from their chair as anything they could ever want could be summoned by the touch of a screen.
ReplyDelete