The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Mark cuban: the world's first trillionaire will be an AI entrepreneur
Poast new message in this thread
Date: February 23rd, 2018 5:46 PM Author: bespoke telephone https://www.cnbc.com/2017/03/13/mark-cuban-the-worlds-first-trillionaire-will-be-an-ai-entrepreneur.html
i dont think the next AI trillionaire though is going to be the person who invents one groundbreaking thing and then just coasts on that for the rest of his life. i think its going to be a person who can take all of the AI technology and synthesize it to pull it all together and make something incredible. i think the next AI entrepreneur who's going to really make it big is going to be a big picture guy.
i'm finding already that so many of these PHD guys who are into machine learning are so narrowly focused. they are very good and are very knowledgeable on one particular subset of machine learning, but when i ask them other questions that don't directly relate to their specific area i find they clam up and just can't answer. that's why i think the person who's going to make it big in this field is going to be a big picture thinker who can synthesize the entire AI ecosystem and communicate it effectively, and not necessarily someone who understands on a very deep level AI theory and exactly HOW the different algorithms work |
Date: February 23rd, 2018 5:48 PM Author: Awkward titillating house friendly grandma i think this guy has more important things to worry about right now (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35472431) |
 |
Date: February 23rd, 2018 5:49 PM Author: amethyst base legal warrant (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35472443) |
|
Date: February 23rd, 2018 5:49 PM Author: shimmering double fault (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35472448) |
|
Date: February 23rd, 2018 6:11 PM Author: Rose Heady Office Son Of Senegal (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35472636) |
|
Date: February 23rd, 2018 6:13 PM Author: bossy chocolate nibblets half-breed The tanking stuff? (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35472653) |
|
Date: February 23rd, 2018 6:58 PM Author: Painfully honest quadroon hospital Lmao no. He was #MeToo’ed (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35473082) |
|
Date: February 24th, 2018 5:50 PM Author: bespoke telephone link? (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479311) |
|
Date: February 23rd, 2018 7:03 PM Author: Well-lubricated church building (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35473115) |
|
Date: February 24th, 2018 7:50 PM Author: angry arrogant rigor (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479980) |
|
Date: February 23rd, 2018 6:18 PM Author: bespoke telephone he is probably right though whatever you think of him (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35472692) |
|
Date: February 23rd, 2018 6:18 PM Author: clear piazza (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35472695) |
|
Date: February 23rd, 2018 6:20 PM Author: shimmering double fault hopefully China steals the tech, combines that with Crispr superhuman operators, and puts us in the dirt (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35472702) |
|
Date: February 24th, 2018 10:23 PM Author: Sadistic location ceo (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35480981) |
|
Date: February 23rd, 2018 6:22 PM Author: know-it-all thriller property electric furnace (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35472720) |
|
Date: February 23rd, 2018 6:55 PM Author: bespoke telephone who? (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35473053) |
|
Date: February 23rd, 2018 7:03 PM Author: Well-lubricated church building (((they))) (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35473117) |
|
Date: February 24th, 2018 4:03 PM Author: bespoke telephone the rothschilds? (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35478763) |
|
Date: February 24th, 2018 6:32 PM Author: know-it-all thriller property electric furnace (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479554) |
|
Date: February 23rd, 2018 6:56 PM Author: Claret stead why would i listen to this guy? he got #metoo'ed (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35473061) |
|
Date: February 24th, 2018 4:20 PM Author: Comical church twinkling uncleanness Herp derp whoever really brings it all together and does something really awesome will make a lot of money, I'm so insightful (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35478831) |
|
Date: February 24th, 2018 4:30 PM Author: bespoke telephone haha touche
no but what i'm really getting at is the guy who can see the big picture, and who isn't necessarily a technical expert at all of the permutations AI has taken, will be the winner after this all shakes out
who can somehow connect what's being done with natural language processing with neural networks with another subset of AI, as an example
because right now, AI seems very segmented, and many AI "experts" are only experts in a very narrow band of AI. those types of people will be needed, obviously, but will not be the true winners
i don't even think the winner in AI will have a PHD in the subject area |
|
Date: February 24th, 2018 5:54 PM Author: Painfully honest quadroon hospital If you had to pick a person that would be able to accomplish this, who would it be? (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479327) |
|
Date: February 24th, 2018 5:59 PM Author: clear piazza Someone at Google (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479357) |
|
Date: February 24th, 2018 6:00 PM Author: bespoke telephone not sure really (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479368) |
|
Date: February 24th, 2018 6:41 PM Author: Awkward titillating house friendly grandma Michael O. Church (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479582) |
|
Date: February 24th, 2018 10:22 PM Author: House-broken hairy legs degenerate (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35480971) |
|
Date: February 24th, 2018 6:24 PM Author: bateful outnumbered cuckold ai solves narrow clearly defined problems, why do you think these things can be linked together (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479505) |
|
Date: February 24th, 2018 6:31 PM Author: ebony headpube lay recurrent neural networks are general computers. there has to be some training algorithm to get them to replicate the functionality of the brain. (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479553) |
|
Date: February 24th, 2018 7:49 PM Author: sticky tank boltzmann why do you think this is true (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479974) |
|
Date: February 24th, 2018 8:14 PM Author: ebony headpube lay Because it is quite unlikely the brain is a hypercomputer. All existing evidence supports the view that it can be simulated as a normal computational process. This is a pretty consensus view in neuroscience and CS. (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35480174) |
|
Date: February 24th, 2018 10:57 PM Author: charcoal gas station the brain isn't a hypercomputer but it's really big and complicated. (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35481203) |
|
Date: February 25th, 2018 2:36 PM Author: ebony headpube lay I think the complexity of the brain is not as high as most people imagine.
http://lesswrong.com/lw/md2/the_brain_as_a_universal_learning_machine/
The cortex is highly regular. Additionally, the problems that AI has making rapid progress on over the last few years are likely some of the hardest - vision and auditory process have undergone much more evolutionary refinement than higher level abstract and logical reasoning. From this perspective, the fact we are seeing superhuman performance in tasks like image recognition, facial recognition and speech processing should be very alarming. It is becoming increasingly difficult for me to imagine we won't see AGI in the next 30 years. |
|
Date: February 24th, 2018 6:05 PM Author: ebony headpube lay AGI is much more likely to come out of a large company like Google than be created by an entrepreneur who has some creative insight into the problem. The reason is that progress is very closely related to available computing power. Efforts like DeepMind/Google Brain have an enormous edge with their ability to rapidly iterate through ML architectutes to find ones that work. Deep reinforcement learning is tricky and requires the optimization of many different parameters to get a working agent. This is basically done through trial and error.
The way ML research appears to be going - learning the learning algorithms themselves - seems to provide an even stronger advantage to established companies that can provide huge computing budgets to their AI research efforts. |
|
Date: February 24th, 2018 6:07 PM Author: clear piazza cr see my poast above it will be Google |
|
Date: February 24th, 2018 6:12 PM Author: ebony headpube lay I would put a 50-60% chance on it being Google the way things are going. None of the other U.S. tech companies seem to be making an effort even close to Google in RL research. I suspect it will either be them or some group in China. (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479434) |
|
Date: February 24th, 2018 6:15 PM Author: Cerebral bipolar box office factory reset button Probably whoever first gets ahold of a working quantum computer (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479449) |
|
Date: February 24th, 2018 6:22 PM Author: ebony headpube lay As far as I can tell there isn't much reason to believe that quantum computers would help much with ML research. Memory bandwidth would likely be a problem, and it is not clear they will help solve the sort of optimization problems we need solved.
Specialized architectures like Graphcore or Google's TPU will greatly speed up research. Eventually neuromorphic computing systems could provide enormous speed and power advantages. |
|
Date: February 24th, 2018 8:08 PM Author: Cerebral bipolar box office factory reset button hmm (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35480127) |
|
Date: February 24th, 2018 6:55 PM Author: bespoke telephone whats your background in all this? curious
lets chat offline if you have an email. could we exchange? |
|
Date: February 24th, 2018 7:11 PM Author: bespoke telephone what type of things could google use and then sell to customers if they develop this first? (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479740) |
|
Date: February 24th, 2018 7:30 PM Author: ebony headpube lay Pretty much anything. I believe hard takeoff is quite likely now when AGI is created. If someone had been able to train an AGI 10 years ago, it would have taken a long time for it to acquire human level expertise in many domains. given current hardware, there will likely be rapid progress to superhuman performance across many important areas. Imagine AlphaZero type learning curves for biomedical research, semiconductor design, corporate strategy, even things like generative models for movies, etc. they could easily dominate any industry they felt like. (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479830) |
|
Date: February 24th, 2018 7:35 PM Author: bespoke telephone that's the answer i was expecting (and afraid you were going to say)
can you imagine the implications of this though? google or whoever develops this will be able to sell this to the highest bidder. think of it - a tool for, just as an example, providing the best corporate strategy for your business to dominate your entire industry and destroy your competition. companies will be fighting over that tool. companies would KILL for that tool.
AGI is scary as fuck to me, because it will be such an incredibly powerful tool. it will be the last invention of the human race. |
|
Date: February 24th, 2018 7:43 PM Author: ebony headpube lay I am deeply concerned about how it will be used. My suspicion is that this will end very badly for humanity, but I hope I am wrong. (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479938) |
|
Date: February 24th, 2018 7:44 PM Author: bespoke telephone dude
lets talk more about this offline
whats your email?
i'm glad you have my concern. these tools are very powerful, but we must be careful. |
|
Date: February 24th, 2018 7:49 PM Author: boyish maize parlor Ok, your AI machine tells your company that you have to acquire X. You (the last human at the company), then go to X, and tell them that your computer told you that you needed to buy X. The guy at X (he's the last one too), asks his computer what he should do to maximize the value of X, and his computer tells him not to sell X to you, but that he should instead go expand into business line Z. So he tells you (via email) to go fuck yourself.
What's so scary about that? |
|
Date: February 24th, 2018 8:31 PM Author: ebony headpube lay My personal view is that keeping humans in the loop as decision makers is not a stable solution in a world with AGI. If someone wants to use an AI consultant in their company to offer guidance (but doesn't give it freedom to act), they will be rapidly outcompeted by another company where there is no human to slow or weaken decision making. The same goes for countries and their militaries.
This means that people will be programming all sorts of utility functions (many of them stupid or contrary to most of humanity's welfare) into superhuman reinforcement learners that have the power to take actions in the real world. In some sense this isn't too different from what we have now (people often have opposing goals). My concern is that utility functions that conflict directly with human welfare could easily have an advantage here, and outcompete the agents we actually want.
The winner take all scenario is also potentially bad but for different reasons. |
|
Date: February 24th, 2018 8:53 PM Author: boyish maize parlor Your outlook is just too negative, IMO. And, even if you end up being right, this is all going to be too big to stop. Why worry about it? Invest in crypto and let do. (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35480468) |
|
Date: February 24th, 2018 10:17 PM Author: ebony headpube lay I hope i am wrong. Properly used AI could make the world a much better place. (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35480927) |
|
Date: February 25th, 2018 12:42 PM Author: bespoke telephone agreed (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35483418) |
|
Date: February 24th, 2018 8:56 PM Author: bespoke telephone how realistic is that winner take all scenario though?
it would seem to me that once AGI is created, it is just going to be a matter of time before a few players dominate every single industry. there will basically be no room for new entrants in any given industry. competition will be eradicated.
i'm trying not to be pessimistic here, though. but it is scary. it's like new entrants to the market will be going against a fucking goliath. it will be like gary kasparov playing against deep blue. except this time it won't be a chess game. |
|
Date: February 24th, 2018 7:39 PM Author: bespoke telephone whoever invents AGI will have almost total control of the fucking world it seems.
that is such a scary fucking concept. can you even fucking imagine?
artificial intelligence fascinates me endlessly. but it also terrifies me, because if we're not careful this thing is going to get way out of hand. |
|
Date: February 24th, 2018 6:13 PM Author: Cerebral bipolar box office factory reset button BUY MATRIX AI NETWORK TOKEN (MAN) (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479439) |
|
Date: February 24th, 2018 7:41 PM Author: boyish maize parlor That's so crazy. Just crazy to think about. I can't believe he thought that and expressed it in an interview like that, but I bet he's right. What a wonderful world we live in. Just wonderful. (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35479919) |
|
Date: February 24th, 2018 10:26 PM Author: Sadistic location ceo Eh talk to me in 30 years (http://www.autoadmit.com/thread.php?thread_id=3902015&forum_id=2#35481011) |