Funnily enough, that segment on Sam Altman is as germane as ever with the release of o1 today. OpenAI reports it as performing as well as PhD students and at least from initial impressions, it appears to be a big step up from GPT-4. As the gaps between significant release dates seems to be decreasing with the rate of progress seemingly accelerating, it feels as if the opportunity cost of unmachining is only rising. Altman's recent comments about strapping rockets to dumpsters don't engender much faith in anyone who wants to believe he is taking the risks seriously.
I just can't help but feel that immediate future of technology will only leave more people behind as opposed to taking them forwards.
Funny kind of PhD students? The mind-set goes round in circles? Not sure Britain pays much attention to PhD graduates anyway, and I am not sure where the clamjamfry of algorithms is taking us, but we are tied to the commercialisation of American invention, as ever.
Don't know much about Altman actually, but first met 'transhumanists' in early incarnation online 1996. Seriously weird ... exteme individualism intent on self-preservation and belief. They even had a political party and a candidate for POTUS. I suppose they were probably an outgrowth of the earlier technocrats? I gather their inspiration is being tied to rockets ... lots and lots of rockets ... oh dearie me.
Goodness, I am only as far as Sam Altman and his consciousness and limited imagination. The rest of your good conversation could take some time.😊
It might be useful to categorise AI. The recent realisation of Artificial General Intelligence seems mostly to be impacting via consumer electronics specifically it appears where AGI exerts a recursive training effect on the human mind via existing networks such as advertising, social media, education, entertainment, access to services and crime? Scientific or technical or legal investigations? ... I suppose there is extensive summarising of 'literature searches'? I guess actual direct use of AI in investigative tools in technical and industrial, financial, medical or military environments ('data handling') will be viaproprietary specific tools, already legion in devices used perhaps more in industrial than consumer applications? Will you wear a medical device?
Attitudes? The other day I read a transcript of a conversation between two good American medical scientists, Eric Topol and Francis Collins. The emphasis inevitably was on the good things to come despite the profound and lasting shock they both had experienced in the handling of covid in the US.
That question posed to Altman rings a signal bell. I can go back to my realising the hubristic trajectory one early summer afternoon in 2006 outside the then new HQ of the Royal Bank of Scotland. The bank had taken advantage of digital communication to expand to a global entity. I could sense the gigantic data streams from machine to machine across the global network. The bank was not to go bust until the contagion of the financial crash caught them in the gap between algorithms and reality.
The growing demand for resources from electronics, especially by data handling and its access via AI is seeing, pretty much by definition unsustainable growth, including the demand for electricity. What the world might be left with is a good question I do not have answers to, and the machines don't understand anything.
I think the issue is that narrow AI, or task-specific AI, is quickly becoming obsolete or subsumed by AGI. There's no reason why a self-driving car couldn't be able to host a multimodal LLM. Obviously there is the question of whether the car should do that, but the point is that AGI would be ubiquitous.
That we should endure energy or resource scarcity seems unlikely given the amount of money it's making, not to mention saving as well by increasing productivity. It's an arms race where business are willing to spend big for access to a responsive and effective model. They'll foot the bill. With how efficient models are now with respect to their quality of output, I don't see AGI needing a Dyson sphere to come online. The entirety of the internet is available for the scraping.
I guess you are right about the arms race dynamic which includes the military arms race for energy and materials.There are some well-developed scenarios about the need for energy to underpin the industrial economies. And the system grows.
Digitalisation grew in tandem with global growth. It is not about efficiency it seems; AI on top spurs/needs growth of the total economy? Imagine the American economy, notoriously profligate relative to the rest of the industrial world in its use of energy, doubles in size in 25 years (3% growth rate) and requires even approximately to be maintainaned at that level of energy and materials input, let alone have the dynamic continue to drive the trajectory. Well, that is exactly what these guys envisage it seems ... 'round the bend' it seems if we look at the numbers I find myself looking at.
And that was just talking about AI development in the U.S., not even mentioning the rest of the globe. With countries like China, there is certainly the concern of making an AGI in their image, one would that implicitly seek to enforce and maintain an authoritarian regime. Even if the U.S. agrees to be more cautious and prudent in their efforts, will everyone else?
One thing that really irks me about the level of dependance upon the Machine-technology that is SUPPOSED to be helpful--is that quite often it may be convenient for the government, but it may also lack the true help that is needed. In our state/county if we are going to dig below a few inches of soil we need to call 811 and make an appointment for the area to be analyzed by the appropriate piece of equipment (?) so we can be assured of not hitting or digging up utility lines. Good idea!! Yes! I dont want to make that mistake! But here is the catch: I had never use the service (811) before and did not know what to expect or how it worked. I understood it to be my only option. So, I called 811 expecting to talk to a live person. NOPE!! No living person to talk to even when I hit every option I could that the system offered. I wanted to know when they would show up. I wanted to know where exactly they would look for lines. How would they know I wanted them to be sure to check the back art of the yard? Were they going to do the WHOLE yard? or just part of it? What if they missed where I wanted to dig? CAN SOMEBODY JUST PLEASE ANSWER THE DUMB PHONE! All I was allowed to do is to make a appointment request via pushing a button, good grief. So, finally I called one of the individual utility companies which I knew would need to look into my digging project. I asked about the specific part of my yard in question. He literally told me," Im sorry but we cannot come onto a persons private property." "Then how am I supposed to know if there are any utility lines underneath," I asked? "Well, you will have to hire a private company to do that," was his answer. Really? Finally, I gave up on speaking to anyone. Some utility people arrived unbeknownst to me, which was a total waste of time for them because I didnt have their services anyhow. They sent me a form indicating that digging was not going to endanger their lines, because, low and behold, I didnt have their services. I could have told them that. This story is getting way too long. But you get the idea of how I feel about automated systems that are supposed to make everything easier and more convenient...LOL. I felt so annoyed! And speaking of human-ness, I think I lost mine for a few moments! Next time Im not calling 811; I think I will call the county and find a real person to speak with, now that I know we can.
Matt crawford wrote something about this here. Basically, many technologies create friction in a system that was once pretty straightforward and did not clearly need any sort of fix. It is brought on for its own sake. It creates friction in a previously smooth process. Friction allows for tech companies to collect newly discovered rents. Pretty clever. Heres that piece:
I have been thinking a lot about the unexamined life vs. the examined life lately. Nadya Williams wrote a piece recently encouraging parents to ask more questions and it strikes me that AI may be deadening our curiosity. If we can "know" anything through an internet search then why would it ever occur to us to ask, as you do, either whether this is reality or whether we want to conform to it?
It is also increasingly difficult to evaluate information to figure out whether it is accurate.
Great and insightful interview.
Funnily enough, that segment on Sam Altman is as germane as ever with the release of o1 today. OpenAI reports it as performing as well as PhD students and at least from initial impressions, it appears to be a big step up from GPT-4. As the gaps between significant release dates seems to be decreasing with the rate of progress seemingly accelerating, it feels as if the opportunity cost of unmachining is only rising. Altman's recent comments about strapping rockets to dumpsters don't engender much faith in anyone who wants to believe he is taking the risks seriously.
I just can't help but feel that immediate future of technology will only leave more people behind as opposed to taking them forwards.
Funny kind of PhD students? The mind-set goes round in circles? Not sure Britain pays much attention to PhD graduates anyway, and I am not sure where the clamjamfry of algorithms is taking us, but we are tied to the commercialisation of American invention, as ever.
Don't know much about Altman actually, but first met 'transhumanists' in early incarnation online 1996. Seriously weird ... exteme individualism intent on self-preservation and belief. They even had a political party and a candidate for POTUS. I suppose they were probably an outgrowth of the earlier technocrats? I gather their inspiration is being tied to rockets ... lots and lots of rockets ... oh dearie me.
OK, I bow out ... salute to our hosts for the conversation ... POV matters. I will continue reading the rest of the transcript!👍
Thanks for taking the time to engage with the ideas we discussed Philip!
Thanks Ruth.
best wishes
Goodness, I am only as far as Sam Altman and his consciousness and limited imagination. The rest of your good conversation could take some time.😊
It might be useful to categorise AI. The recent realisation of Artificial General Intelligence seems mostly to be impacting via consumer electronics specifically it appears where AGI exerts a recursive training effect on the human mind via existing networks such as advertising, social media, education, entertainment, access to services and crime? Scientific or technical or legal investigations? ... I suppose there is extensive summarising of 'literature searches'? I guess actual direct use of AI in investigative tools in technical and industrial, financial, medical or military environments ('data handling') will be viaproprietary specific tools, already legion in devices used perhaps more in industrial than consumer applications? Will you wear a medical device?
Attitudes? The other day I read a transcript of a conversation between two good American medical scientists, Eric Topol and Francis Collins. The emphasis inevitably was on the good things to come despite the profound and lasting shock they both had experienced in the handling of covid in the US.
That question posed to Altman rings a signal bell. I can go back to my realising the hubristic trajectory one early summer afternoon in 2006 outside the then new HQ of the Royal Bank of Scotland. The bank had taken advantage of digital communication to expand to a global entity. I could sense the gigantic data streams from machine to machine across the global network. The bank was not to go bust until the contagion of the financial crash caught them in the gap between algorithms and reality.
The growing demand for resources from electronics, especially by data handling and its access via AI is seeing, pretty much by definition unsustainable growth, including the demand for electricity. What the world might be left with is a good question I do not have answers to, and the machines don't understand anything.
I think the issue is that narrow AI, or task-specific AI, is quickly becoming obsolete or subsumed by AGI. There's no reason why a self-driving car couldn't be able to host a multimodal LLM. Obviously there is the question of whether the car should do that, but the point is that AGI would be ubiquitous.
That we should endure energy or resource scarcity seems unlikely given the amount of money it's making, not to mention saving as well by increasing productivity. It's an arms race where business are willing to spend big for access to a responsive and effective model. They'll foot the bill. With how efficient models are now with respect to their quality of output, I don't see AGI needing a Dyson sphere to come online. The entirety of the internet is available for the scraping.
I guess you are right about the arms race dynamic which includes the military arms race for energy and materials.There are some well-developed scenarios about the need for energy to underpin the industrial economies. And the system grows.
Digitalisation grew in tandem with global growth. It is not about efficiency it seems; AI on top spurs/needs growth of the total economy? Imagine the American economy, notoriously profligate relative to the rest of the industrial world in its use of energy, doubles in size in 25 years (3% growth rate) and requires even approximately to be maintainaned at that level of energy and materials input, let alone have the dynamic continue to drive the trajectory. Well, that is exactly what these guys envisage it seems ... 'round the bend' it seems if we look at the numbers I find myself looking at.
And that was just talking about AI development in the U.S., not even mentioning the rest of the globe. With countries like China, there is certainly the concern of making an AGI in their image, one would that implicitly seek to enforce and maintain an authoritarian regime. Even if the U.S. agrees to be more cautious and prudent in their efforts, will everyone else?
One thing that really irks me about the level of dependance upon the Machine-technology that is SUPPOSED to be helpful--is that quite often it may be convenient for the government, but it may also lack the true help that is needed. In our state/county if we are going to dig below a few inches of soil we need to call 811 and make an appointment for the area to be analyzed by the appropriate piece of equipment (?) so we can be assured of not hitting or digging up utility lines. Good idea!! Yes! I dont want to make that mistake! But here is the catch: I had never use the service (811) before and did not know what to expect or how it worked. I understood it to be my only option. So, I called 811 expecting to talk to a live person. NOPE!! No living person to talk to even when I hit every option I could that the system offered. I wanted to know when they would show up. I wanted to know where exactly they would look for lines. How would they know I wanted them to be sure to check the back art of the yard? Were they going to do the WHOLE yard? or just part of it? What if they missed where I wanted to dig? CAN SOMEBODY JUST PLEASE ANSWER THE DUMB PHONE! All I was allowed to do is to make a appointment request via pushing a button, good grief. So, finally I called one of the individual utility companies which I knew would need to look into my digging project. I asked about the specific part of my yard in question. He literally told me," Im sorry but we cannot come onto a persons private property." "Then how am I supposed to know if there are any utility lines underneath," I asked? "Well, you will have to hire a private company to do that," was his answer. Really? Finally, I gave up on speaking to anyone. Some utility people arrived unbeknownst to me, which was a total waste of time for them because I didnt have their services anyhow. They sent me a form indicating that digging was not going to endanger their lines, because, low and behold, I didnt have their services. I could have told them that. This story is getting way too long. But you get the idea of how I feel about automated systems that are supposed to make everything easier and more convenient...LOL. I felt so annoyed! And speaking of human-ness, I think I lost mine for a few moments! Next time Im not calling 811; I think I will call the county and find a real person to speak with, now that I know we can.
Matt crawford wrote something about this here. Basically, many technologies create friction in a system that was once pretty straightforward and did not clearly need any sort of fix. It is brought on for its own sake. It creates friction in a previously smooth process. Friction allows for tech companies to collect newly discovered rents. Pretty clever. Heres that piece:
https://mcrawford.substack.com/p/what-is-tech
Thank you for this interview!
I have been thinking a lot about the unexamined life vs. the examined life lately. Nadya Williams wrote a piece recently encouraging parents to ask more questions and it strikes me that AI may be deadening our curiosity. If we can "know" anything through an internet search then why would it ever occur to us to ask, as you do, either whether this is reality or whether we want to conform to it?
It is also increasingly difficult to evaluate information to figure out whether it is accurate.