The most prestigious law school admissions discussion board in the world.
|
 |
 
|
OpenAI Codex… Tech Unemployment Forthcoming?
| glittery party of the first part den | 11/27/21 |
| avocado church | 11/27/21 |
| glittery party of the first part den | 11/27/21 |
| Aquamarine new version trust fund | 11/27/21 |
| cordovan principal's office friendly grandma | 11/27/21 |
| Aquamarine new version trust fund | 11/27/21 |
| cordovan principal's office friendly grandma | 11/27/21 |
| Costumed Bearded Halford Roommate | 11/27/21 |
| narrow-minded aqua state | 11/27/21 |
| iridescent provocative national security agency sneaky criminal | 11/27/21 |
| bat-shit-crazy titillating depressive wrinkle | 11/27/21 |
| fertile young woman plugged into zogbox 10 hrs/day | 07/08/26 |
| supple hunting ground | 11/27/21 |
| gaetan dugas | 07/08/26 |
| OYT and the Indie Reprieve | 07/08/26 |
Poast new message in this thread
Date: November 27th, 2021 4:34 PM Author: glittery party of the first part den (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#43518127) |
Date: November 27th, 2021 4:34 PM Author: avocado church not even close (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#43518131) |
 |
Date: November 27th, 2021 4:40 PM Author: glittery party of the first part den Why not? (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#43518151) |
|
Date: November 27th, 2021 4:44 PM Author: Aquamarine new version trust fund https://arxiv.org/abs/2111.08267 https://arxiv.org/abs/2111.08171
So basically with a little prompt engineering Codex can solve college level probability and linear algebra problems. Codex won't directly or immediately cause tech unemployment. that's not the point anyway. the implication of it is that the neural network scaling hypothesis is true and that NNs, when of sufficiently large size and trained on a wide range of experiences, start meta-learning. no need to study the brain in detail to figure out what the neocortex is doing and how to replicate it in silicon - just train NNs on large datasets and the neural networks will meta-learn powerful algorithms that generalize.
timelines to AGI are increasingly short. |
|
Date: November 27th, 2021 5:03 PM Author: cordovan principal's office friendly grandma I fail to see how the implication you draw follows. (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#43518251) |
|
Date: November 27th, 2021 5:16 PM Author: Aquamarine new version trust fund it's just repurposed GPT-3, fine-tuned for code generation on another dataset. the code generation capabilities are one part of the picture here. the same model has very impressive language comprehension and generation capabilities considering its architectural simplicity. the coding capabilities are more evidence that the model can be pushed to do many things with appropriate prompt engineering.
not to mention transformers are also work for image generation, reinforcement learning, image recognition, etc. if we were having to create highly specific models to solve certain tasks, that would be evidence in favor of the view that AGI will require substantial engineering effort. we are seeing the exact opposite. even a simple highly regularized MLP can accomplish what a transformer can when trained on enough data. |
|
Date: November 27th, 2021 5:28 PM Author: cordovan principal's office friendly grandma Your conclusion seems to be about "meta-learning" from "large datasets", presumably leading to some sort of "foom" scenario, but the examples you've given are of narrowly-circumscribed problems for which there are large datasets available. So, I am having a hard time following how your conclusion follows from your premises. Wouldn't the more modest conclusion be that, in areas with sufficient data and amenable to production of large amounts of data and which are otherwise tractable, AI will prove useful? I don't see any hint of universal generalization.
What is the missing piece that's been filled in that answers questions like those raised by Pearl http://ftp.cs.ucla.edu/pub/stat_ser/r475.pdf
Jordan https://hdsr.mitpress.mit.edu/pub/wot7mkc1/release/9
or Hanson https://spectrum.ieee.org/economics-of-the-singularity |
|
Date: November 27th, 2021 9:40 PM Author: Costumed Bearded Halford Roommate (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#43519775) |
|
Date: November 27th, 2021 8:27 PM Author: narrow-minded aqua state (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#43519385) |
|
Date: November 27th, 2021 4:35 PM Author: iridescent provocative national security agency sneaky criminal highest number of H1B visas in decades. odd case (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#43518134) |
|
Date: November 27th, 2021 5:08 PM Author: bat-shit-crazy titillating depressive wrinkle The ideal situation for tech companies is high supply and low demand. High number of H1Bs doesn't mean demand won't plummet (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#43518276) |
|
Date: July 8th, 2026 3:34 PM Author: fertile young woman plugged into zogbox 10 hrs/day (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#49986578) |
|
Date: November 27th, 2021 9:37 PM Author: supple hunting ground i've heard this shit often gives retarded suggestions even for javascript which you would think is a language with a lot of examples to learn from (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#43519765) |
|
Date: July 8th, 2026 3:28 PM Author: gaetan dugas (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#49986559) |
|
Date: July 8th, 2026 3:30 PM Author: OYT and the Indie Reprieve ( ) lol (http://www.autoadmit.com/thread.php?thread_id=4973764&forum_id=2most#49986564) |