Artificial intelligence won’t end civilization, argues Marc Andreessen. Just the opposite. It is quite possibly the best thing human beings have ever created.
I'm coming to this late and so probably already missed the conversation. I mostly agree with Marc but he neglected pointing out some things. Part of his positives is that it makes stupid people smarter. I agree that is a positive but it's also a negative and here is why. Stupid people are more likely to commit antisocial behavior. What I mean by that is not saying mean things but rather physically trying to hurt other people. As Marc says, AI is just a tool. It has no will, no motivation no ethics. It does whatever the smart or stupid person asks of it and so guard rails have to be put in place to prevent, for example, someone asking the AI to find the most deadly chemical compounds, then asking the ai how to secure the ingredients and build a small factory to produce them. The same power AI will give researchers, artist, engineers, etc. it gives to radicals, warmongers, cultist and crazy people that just want to see the world burn.
So yes a lot of what Marc foresees will happen. Crunching through every possible protein fold and then highlighting the most promising for lets say leukemia treatment, is something computers are very good at. It's going to get easier to make libraries of chemistry, biology and physics. This will improve life, energy recourses, materials, water treatment, education and the list goes on. That same flame that warms your hand can burn you though and it's like the nuclear bomb or crisper. We are entering a new epic. The things that will keep you up at night will increase along with the things we no longer have to worry about like improved energy, water treatment and food recourses. Ai will help us solve a lot of world problems but it will also introduce new problems. Such is life!
My simple mind has one question: considering that humans create AI and humans are flawed, imperfect and changing in nature, how AI is not at risk of being flawed itself?
AI will be self improving, ok on what basis? To my knowledge when we identify that we fall short of an expectation and we have a path to reach the outcome.
The worst historic periods have started on the best intentions of a fairer better society and ended in atrocities for humans.
So goals can be virtuous in theory and create terrible side effects in practice. If AI reaches a point where it masters the understanding of human unstable nature. AI will be very dangerous indeed.
I am a firm believer that progress and creativity can solve our problems. I am not sure all problems ought to be solved though. I have lived on 3 different continents and I am struck by how different we are in terms of values, sense of purpose and happiness, etc... it is naive or dogmatic to think we all want the same things. Therefore AI will naturally serve quicker and better a dictatorship that decides for all humans than serve an infinite amount of different aspirations or views that are moving targets.
Moreover the warfare comment stunned me. And I wonder if I am completely dumb or if this highly intelligent successful man is ... naive.
Wars are about expansion, ressources acquisition, dogmatisme, destruction. AI into looks to me like nuclear weapon on steroids. Sure it will make this shorter by orchestrating mutual destruction faster. And that is great news?!
Last AI panic as he calls it is more about the pace of this change than the change itself. Our brains need to process and we all know that no one processes information at the same pace. When we don’t have time to process new information we back off and fear it. It is a natural human reaction. And please note that we are most likely the fastest adaptable living creature.
So AI is not to me something not to be thoroughly discussed and understood before implementation.
Leaving a few ... men decide what AI is going to based on seems to me a very risky bet for us all.
When AI self corrects how will we be able to change the track if we think it is wrong or realize afterwards that is dangerous ?
This essay was depressing and did nothing to allay my concerns over AI. Nowhere did it touch on what makes us human. Marc A seems to view AI as optimization for optimization's sake while failing to ask the deeper questions about the what and the why. Where is the humanity, or the joy in that? Perhaps the question is not really for technologists (or venture capitalists) to try and answer in any case. Perhaps it's a question for philosophers. At the least I'd be better comforted if I was sure it would ultimately be answered by a fully realized human without the crutch of, or submission to, a machine.
I can’t get past the infinitely kind and patient therapist/tutor/coach etc. My kids’ classmates are already having problems with social cues, facial expressions, etc. Isn’t part of being social creatures learning the subtle and complex signals of others? We don’t live in a world of infinitely patient and perhaps should not - lest our future children never learn when they have pushed passed the right use of another for assistance and into detrimental reliance. This and other examples of the author’s exuberance give me shivers as we move so far beyond being human - what it means and what it is. I picked up Sapiens again for a re-read after this article.
We have lost a generation to the brain suckers and they are coming for the rest of us. Now interests will be better able, in the interest of profits and power and without restraint of good will and good governance, to exploit our addictive itches. Mammon will take form as a station on the road to a higher consciousness of non-human values. See essay ‘The Man in the Machine’ and story ‘Saucerville’ on my posts.
It won't end civilization any more than cars ended transportation. As Gilder said a few decades back, the incredible thesis of many is wealth is the cause of poverty, water is the cause of thirst, food is the cause of hunger. I will not thrash against advances that might be misused (like guns or drugs or cars) because I am fearful of the misuse. I am not trading in my cell phone and laptop to raise sheep and potatoes in Ireland. I embrace progress and the conveniences our technology and relative wealth provides. Poverty is down 90% in the last century because of progress. There is NO evidence of some grand plan, no matter how the cards are stacked or Birchers like charts are constructed. The hand wringing is overwrought and misguided. Are there risks? Sure, just like the risk I take that my car will not burn me alive when I turn the key. All of life is a trade off, you pick the trades you want to make.
This is possibly the worst-reasoned article I've ever seen on CS/TFP, and they have had a few doozies.
According to Andressen, AI will be infinitely compassionate, helpful…it will provide infinite love.
Also according to him, AI is just a technology like fire or the telephone, nothing to get excited about. It’s just math and code.
You can’t have it both ways. If AI actually provides *compassion* and *infinite love* and companionship, as he claims, we’re talking about something very different from math and code. But if it is just math and code, the companionship, love and compassion are merely illusions. Pick one, Andreesen, but you can’t have both.
The business about Baptists, and California cults was truly pathetic. It was a blatant, but rambling and largely incoherent attempt to imply that those who disagree with him are just nutty religious zealots. Yet, he utterly failed to make that case in a consistent, rational way. That part kind of sounded like it was written when he was high. By the time he was done with the article, HE was the one who sounded like a zealous, irrational, true believer.
For a moment I was grateful for his brief recap of the benefits of a free market economy, which was succinct, and largely correct. But he can’t have it both ways. He began the article by stating that AI is like nothing that every came before, but then he says, “Don’t panic, it’s just like every other technological innovation that came before.” So, is AI like every other technology, or is it unique? You can’t have it both ways.
I am disappointed, TFP. This is not thoughtful discourse. Maybe AI will be relief, in that it surely has to be more logical than this.
It feels like this essay was written by an AI chatbot. It would help in the intro to have a better sense of who Marc Andreessen is besides just a "venture capitalist." I suspect he's someone who has a vested interest in the technology, as his pro-AI manifesto clearly comes across that way. I just skimmed the article after the list that began with the way every kid will have an "infinitely loving," etc., personal AI tutor. (Yeah that'll happen.) It's unreadable. Bari, with these articles coming in every day, the average subscriber who doesn't have unlimited leisure time can't read something of this length. Some serious editing and streamlining would help immensely.
Mr Andreessen makes no effort to persuade the reader who is skeptical of AI. Instead, he writes to dismiss the skeptic as a conspiracy theorist or an ignoramus. His article drips with condescension. It's clear that he has not tried to learn about the real reasons people fear AI. For example, something being "math and code" made by humans and not "alive" does not mean it is incapable of having goals. The fact that he brushes aside the argument that AI could develop goals (or just take very seriously the goals given by humans, in a way unconstrained by human sense and morality - see Bostrom's paperclip maximizer thought experiment) with "but it's not alive" shows just how ignorant he is of the threat we face with AI. He clearly enjoyed knocking down the strawmen he built. I also laughed when he wrote that coastal elites do not represent humanity and don't get to decide what happens... Folks who don't know who Marc Andreessen is need to look him up. Pot calling the kettle black! And that's not to even mention his immense conflict of interest as a venture capitalist backing tech firms who develop AI. Take this essay with a planet-sized grain of salt.
I still think AI tutors for students and AI coaches / mentors etc will always lack the human wisdom gained from life experiences, successes and failures. Not convinced AI can ever capture (or teach) that. Intelligence isn’t everything. I know lots of intelligent people who lack common sense, wisdom and morality. Those things are passed on from fellow humans.
My qualifications for commenting on Andreessen's article, besides being a "grumpy old mansplainer" (as my profile says):
I'm not an expert on AI, but I have been programming computers since 1962 (!), with the last several decades focused on predictive analytics (and its related areas: data mining, machine learning and forecasting, pattern recognition, etc.). Although long retired from my day job, I still practice my skills in machine learning contests, including some that involve "deep learning" of the kind used to generate those large language models like ChatGPT, although on a much smaller scale.
I can't help but cringe when Andreessen says things like: "Every child will have an AI tutor that is infinitely patient, infinitely compassionate, infinitely knowledgeable, infinitely helpful.". I agree that AI tutors may be quite useful if properly designed, but to rhapsodize about the humanizing potential of AI the way Andreessen does makes me wonder whether he might be under the spell of a secret robot lover. The word "compassionate" implies having emotions and motives, neither of which (as far as I know) underpin any of today's AI contraptions, and that explains why ChatGPT can sound fairly rational for a while and then go completely off the rails into gibbering nonsense. I think that ChatGPT and other large language models are basically still "fancy auto complete" mechanisms. I am actually a "strong AI" "optimist" who believes that AI might eventually acquire some version of the human attributes of emotions and motives, in which case it may seriously threaten to pass the "Turing test", but that's as yet at some indefinite time in the future. Andreessen might say that for practical purposes it's enough for AI to convincingly fake having emotions like compassion, but I doubt that will be possible without having something resembling the real deal at its core.
Andreessen admits that "AI will make it easier for bad people to do bad things", but he also claims that "AI is going to improve warfare", which sounds like (and I believe is) an oxymoron. Long before AI decides to wage war on humanity in Terminator style it will be capable of much destructive mischief. Countries and non-state actors are already using swarms of "intelligent" drones to attack their enemies. Although Andreessen says that the answer is to "mount major efforts to use AI for good, legitimate, defensive purposes", the problem is that offense has a natural advantage over defense, which is often stuck playing an imperfect game of catchup.
Where Andreessen does get some things right is on economic principles, including a good explanation of the "lump of labor fallacy", and why the rising tide of a flexible market economy tends to lift all boats.
Sounds like we'll all have and AI and over time, it will become the same, best, AI and everyone will know the same thing and think the same thing and we'll lose innovation and risk taking.
Many of the utopian ideals Marc projects are actually the death of civilization as we know it. Assistants that are "... infinitely patient, infinitely compassionate, infinitely knowledgeable, infinitely helpful" could be terrible for humanity.
Lack of patience, knowledge, helpfulness, and even compassion are important feedback mechanisms, if properly regulated, in any educational endeavor.
Without these, you end up with a society of entitlement and unreasonable expectations, together with a brittle inability to deal with adversity and challenge.
Bingo!! I would assert that anyone with children born post 2000-2004 already recognizes the affects listed in your last sentence. His utopian ideals about how much intelligence will proliferate is already being disproven by this generation. Soooo much information at SUCH easy access, yet the average person prefers the sedative of easy access to constant indulgence in entertainment over the opportunity to grow in knowledge.
I always wonder about these “tech-savvy” sorts and whether or not they have children! (The ones who do have quite strict boundaries on their children’s usage.)
I'm coming to this late and so probably already missed the conversation. I mostly agree with Marc but he neglected pointing out some things. Part of his positives is that it makes stupid people smarter. I agree that is a positive but it's also a negative and here is why. Stupid people are more likely to commit antisocial behavior. What I mean by that is not saying mean things but rather physically trying to hurt other people. As Marc says, AI is just a tool. It has no will, no motivation no ethics. It does whatever the smart or stupid person asks of it and so guard rails have to be put in place to prevent, for example, someone asking the AI to find the most deadly chemical compounds, then asking the ai how to secure the ingredients and build a small factory to produce them. The same power AI will give researchers, artist, engineers, etc. it gives to radicals, warmongers, cultist and crazy people that just want to see the world burn.
So yes a lot of what Marc foresees will happen. Crunching through every possible protein fold and then highlighting the most promising for lets say leukemia treatment, is something computers are very good at. It's going to get easier to make libraries of chemistry, biology and physics. This will improve life, energy recourses, materials, water treatment, education and the list goes on. That same flame that warms your hand can burn you though and it's like the nuclear bomb or crisper. We are entering a new epic. The things that will keep you up at night will increase along with the things we no longer have to worry about like improved energy, water treatment and food recourses. Ai will help us solve a lot of world problems but it will also introduce new problems. Such is life!
My simple mind has one question: considering that humans create AI and humans are flawed, imperfect and changing in nature, how AI is not at risk of being flawed itself?
AI will be self improving, ok on what basis? To my knowledge when we identify that we fall short of an expectation and we have a path to reach the outcome.
The worst historic periods have started on the best intentions of a fairer better society and ended in atrocities for humans.
So goals can be virtuous in theory and create terrible side effects in practice. If AI reaches a point where it masters the understanding of human unstable nature. AI will be very dangerous indeed.
I am a firm believer that progress and creativity can solve our problems. I am not sure all problems ought to be solved though. I have lived on 3 different continents and I am struck by how different we are in terms of values, sense of purpose and happiness, etc... it is naive or dogmatic to think we all want the same things. Therefore AI will naturally serve quicker and better a dictatorship that decides for all humans than serve an infinite amount of different aspirations or views that are moving targets.
Moreover the warfare comment stunned me. And I wonder if I am completely dumb or if this highly intelligent successful man is ... naive.
Wars are about expansion, ressources acquisition, dogmatisme, destruction. AI into looks to me like nuclear weapon on steroids. Sure it will make this shorter by orchestrating mutual destruction faster. And that is great news?!
Last AI panic as he calls it is more about the pace of this change than the change itself. Our brains need to process and we all know that no one processes information at the same pace. When we don’t have time to process new information we back off and fear it. It is a natural human reaction. And please note that we are most likely the fastest adaptable living creature.
So AI is not to me something not to be thoroughly discussed and understood before implementation.
Leaving a few ... men decide what AI is going to based on seems to me a very risky bet for us all.
When AI self corrects how will we be able to change the track if we think it is wrong or realize afterwards that is dangerous ?
This essay was depressing and did nothing to allay my concerns over AI. Nowhere did it touch on what makes us human. Marc A seems to view AI as optimization for optimization's sake while failing to ask the deeper questions about the what and the why. Where is the humanity, or the joy in that? Perhaps the question is not really for technologists (or venture capitalists) to try and answer in any case. Perhaps it's a question for philosophers. At the least I'd be better comforted if I was sure it would ultimately be answered by a fully realized human without the crutch of, or submission to, a machine.
This author needs to do much more to back up his own claims than to just discredit the risks. This article is highly unconvincing.
I can’t get past the infinitely kind and patient therapist/tutor/coach etc. My kids’ classmates are already having problems with social cues, facial expressions, etc. Isn’t part of being social creatures learning the subtle and complex signals of others? We don’t live in a world of infinitely patient and perhaps should not - lest our future children never learn when they have pushed passed the right use of another for assistance and into detrimental reliance. This and other examples of the author’s exuberance give me shivers as we move so far beyond being human - what it means and what it is. I picked up Sapiens again for a re-read after this article.
We have lost a generation to the brain suckers and they are coming for the rest of us. Now interests will be better able, in the interest of profits and power and without restraint of good will and good governance, to exploit our addictive itches. Mammon will take form as a station on the road to a higher consciousness of non-human values. See essay ‘The Man in the Machine’ and story ‘Saucerville’ on my posts.
It won't end civilization any more than cars ended transportation. As Gilder said a few decades back, the incredible thesis of many is wealth is the cause of poverty, water is the cause of thirst, food is the cause of hunger. I will not thrash against advances that might be misused (like guns or drugs or cars) because I am fearful of the misuse. I am not trading in my cell phone and laptop to raise sheep and potatoes in Ireland. I embrace progress and the conveniences our technology and relative wealth provides. Poverty is down 90% in the last century because of progress. There is NO evidence of some grand plan, no matter how the cards are stacked or Birchers like charts are constructed. The hand wringing is overwrought and misguided. Are there risks? Sure, just like the risk I take that my car will not burn me alive when I turn the key. All of life is a trade off, you pick the trades you want to make.
This is possibly the worst-reasoned article I've ever seen on CS/TFP, and they have had a few doozies.
According to Andressen, AI will be infinitely compassionate, helpful…it will provide infinite love.
Also according to him, AI is just a technology like fire or the telephone, nothing to get excited about. It’s just math and code.
You can’t have it both ways. If AI actually provides *compassion* and *infinite love* and companionship, as he claims, we’re talking about something very different from math and code. But if it is just math and code, the companionship, love and compassion are merely illusions. Pick one, Andreesen, but you can’t have both.
The business about Baptists, and California cults was truly pathetic. It was a blatant, but rambling and largely incoherent attempt to imply that those who disagree with him are just nutty religious zealots. Yet, he utterly failed to make that case in a consistent, rational way. That part kind of sounded like it was written when he was high. By the time he was done with the article, HE was the one who sounded like a zealous, irrational, true believer.
For a moment I was grateful for his brief recap of the benefits of a free market economy, which was succinct, and largely correct. But he can’t have it both ways. He began the article by stating that AI is like nothing that every came before, but then he says, “Don’t panic, it’s just like every other technological innovation that came before.” So, is AI like every other technology, or is it unique? You can’t have it both ways.
I am disappointed, TFP. This is not thoughtful discourse. Maybe AI will be relief, in that it surely has to be more logical than this.
It feels like this essay was written by an AI chatbot. It would help in the intro to have a better sense of who Marc Andreessen is besides just a "venture capitalist." I suspect he's someone who has a vested interest in the technology, as his pro-AI manifesto clearly comes across that way. I just skimmed the article after the list that began with the way every kid will have an "infinitely loving," etc., personal AI tutor. (Yeah that'll happen.) It's unreadable. Bari, with these articles coming in every day, the average subscriber who doesn't have unlimited leisure time can't read something of this length. Some serious editing and streamlining would help immensely.
Will AI kill humanity? No, we’ll just pull the plug, literally
Mr Andreessen makes no effort to persuade the reader who is skeptical of AI. Instead, he writes to dismiss the skeptic as a conspiracy theorist or an ignoramus. His article drips with condescension. It's clear that he has not tried to learn about the real reasons people fear AI. For example, something being "math and code" made by humans and not "alive" does not mean it is incapable of having goals. The fact that he brushes aside the argument that AI could develop goals (or just take very seriously the goals given by humans, in a way unconstrained by human sense and morality - see Bostrom's paperclip maximizer thought experiment) with "but it's not alive" shows just how ignorant he is of the threat we face with AI. He clearly enjoyed knocking down the strawmen he built. I also laughed when he wrote that coastal elites do not represent humanity and don't get to decide what happens... Folks who don't know who Marc Andreessen is need to look him up. Pot calling the kettle black! And that's not to even mention his immense conflict of interest as a venture capitalist backing tech firms who develop AI. Take this essay with a planet-sized grain of salt.
As a fan of The 100, all I could think about while reading this essay was Allie! Heaven help us all! LOL
I still think AI tutors for students and AI coaches / mentors etc will always lack the human wisdom gained from life experiences, successes and failures. Not convinced AI can ever capture (or teach) that. Intelligence isn’t everything. I know lots of intelligent people who lack common sense, wisdom and morality. Those things are passed on from fellow humans.
And this article’s author may be your “case in point.”
My qualifications for commenting on Andreessen's article, besides being a "grumpy old mansplainer" (as my profile says):
I'm not an expert on AI, but I have been programming computers since 1962 (!), with the last several decades focused on predictive analytics (and its related areas: data mining, machine learning and forecasting, pattern recognition, etc.). Although long retired from my day job, I still practice my skills in machine learning contests, including some that involve "deep learning" of the kind used to generate those large language models like ChatGPT, although on a much smaller scale.
I can't help but cringe when Andreessen says things like: "Every child will have an AI tutor that is infinitely patient, infinitely compassionate, infinitely knowledgeable, infinitely helpful.". I agree that AI tutors may be quite useful if properly designed, but to rhapsodize about the humanizing potential of AI the way Andreessen does makes me wonder whether he might be under the spell of a secret robot lover. The word "compassionate" implies having emotions and motives, neither of which (as far as I know) underpin any of today's AI contraptions, and that explains why ChatGPT can sound fairly rational for a while and then go completely off the rails into gibbering nonsense. I think that ChatGPT and other large language models are basically still "fancy auto complete" mechanisms. I am actually a "strong AI" "optimist" who believes that AI might eventually acquire some version of the human attributes of emotions and motives, in which case it may seriously threaten to pass the "Turing test", but that's as yet at some indefinite time in the future. Andreessen might say that for practical purposes it's enough for AI to convincingly fake having emotions like compassion, but I doubt that will be possible without having something resembling the real deal at its core.
Andreessen admits that "AI will make it easier for bad people to do bad things", but he also claims that "AI is going to improve warfare", which sounds like (and I believe is) an oxymoron. Long before AI decides to wage war on humanity in Terminator style it will be capable of much destructive mischief. Countries and non-state actors are already using swarms of "intelligent" drones to attack their enemies. Although Andreessen says that the answer is to "mount major efforts to use AI for good, legitimate, defensive purposes", the problem is that offense has a natural advantage over defense, which is often stuck playing an imperfect game of catchup.
Where Andreessen does get some things right is on economic principles, including a good explanation of the "lump of labor fallacy", and why the rising tide of a flexible market economy tends to lift all boats.
Sounds like we'll all have and AI and over time, it will become the same, best, AI and everyone will know the same thing and think the same thing and we'll lose innovation and risk taking.
Anyway, I'm just waiting for my sex robot.
Many of the utopian ideals Marc projects are actually the death of civilization as we know it. Assistants that are "... infinitely patient, infinitely compassionate, infinitely knowledgeable, infinitely helpful" could be terrible for humanity.
Lack of patience, knowledge, helpfulness, and even compassion are important feedback mechanisms, if properly regulated, in any educational endeavor.
Without these, you end up with a society of entitlement and unreasonable expectations, together with a brittle inability to deal with adversity and challenge.
Bingo!! I would assert that anyone with children born post 2000-2004 already recognizes the affects listed in your last sentence. His utopian ideals about how much intelligence will proliferate is already being disproven by this generation. Soooo much information at SUCH easy access, yet the average person prefers the sedative of easy access to constant indulgence in entertainment over the opportunity to grow in knowledge.
I always wonder about these “tech-savvy” sorts and whether or not they have children! (The ones who do have quite strict boundaries on their children’s usage.)