Searching...
16 results for “ai limitations”
experiences is really powerful. And that also segues into simulation. Simulation is really important both for human experiences, human learning, as well as for embodied AI. Robots needs to learn from simulation as much as it needs to learn from the r
6:45 / 7:46
from transcript6:45 – 7:46
in America and in a number of Western countries is a problem. And in contrast, I see China building platforms left and right, including nuclear, so that will be an interesting dynamic. And then semiconductors is another bottleneck. But AI is so compl
4:21 / 5:31
from transcript4:21 – 5:31
Prompts could be a sentence or an entire archive. Outputs can end instantly or stretch on indefinitely. Thousands of users can arrive at once, each making incompatible demands on the same hardware. And all of this has to happen in real time on GPUs t
1:28 / 2:34
from transcript1:28 – 2:34
You can go and and kinda train it against that, but you can never be certain with any strong degree of accuracy that it won't happen again. This does start to feel like a little bit like the Aliven problem where, like, in theory, you know, it's like
1:16:26 / 1:17:30
from transcript1:16:26 – 1:17:30
using new technology, in this case, general purpose technology and repurposing it for harmful ends. And that's a huge, risk. And I think that has a lot of complications because, generally, you know, I'm in huge favor of open science and open source.
2:07:29 / 2:08:32
from transcript2:07:29 – 2:08:32
so I think there's a good chance that that could happen. Part of it is what what is your definition of AGI? Of course, people are arguing about that now, and and, mine's quite a high bar and always has been of, like, can we match the cognitive functi
59:15 / 1:00:19
from transcript59:15 – 1:00:19
enterprise market, but it is a hard like, I don't know how you get people to try doing that. I guess if ChatGPT can do it, like, people are trying ChatGPT. I think it boils down to the the scientific question of how hard is tool use to solve. Because
3:32:14 / 3:33:28
from transcript3:32:14 – 3:33:28
especially with this nondeterministic API again. Right? So you're kind of stuck because, there are tons of instructions that you could inject within your prompt and then, yeah, it's it's going really bad. Okay. I let's actually spend a little time he
23:15 / 24:18
from transcript23:15 – 24:18
But the things you will see a lot are models being tricked into generating, like, porn, or hate speech or phishing messages or viruses, computer viruses. And these are truly harmful impacts and truly an AI safety slash security problem, but the bigge
56:10 / 57:19
from transcript56:10 – 57:19
but there are other paradigms as well. If they're the ones paying for the inference, is there not less incentive to make them efficient? Because, sod it, you're not the one paying for it. If you're the one paying for the inference costs, god, I want
9:29 / 10:58
from transcript9:29 – 10:58
there is no single right way of doing it. You can decide how to constrain that autonomy. Right? A very I mean, a different example of how you could constrain autonomy is preauthorization use cases. Insurance preauthorization is a very ripe use case f
16:28 / 17:30
from transcript16:28 – 17:30
...soon the limitations of that started becoming apparent. I mean, now the next layer of complexity reveals itself. So that's my view. I I wouldn't put too much stock into these overconfident predictions from CEOs. Is it possible to have a dual strategy
22:18 / 23:31
from transcript22:18 – 23:31
Sometimes we do intellectually really easy tasks. Like if someone asked me, like, yesterday my daughter, she misspelled the word butterfly, so I need to tell her how to spell butterfly. It's a low, you know, it's an easy intellectual task. And someti
33:39 / 34:48
from transcript33:39 – 34:48
...limitations? That's a terrifying place to be as one of these frontier labs and and companies because millions of people are using them. There was a lot of backlash last year with the GPT four o getting removed. And I personally never used the model,
1:38:13 / 1:39:17
from transcript1:38:13 – 1:39:17
a couple of office rooms and ask the the model to count the number of chairs. And this is something a toddler could do or maybe maybe a a elementary school kid could do, and AI could not do that. Right? So, there's just so much AI today could not do.
27:37 / 28:46
from transcript27:37 – 28:46
We've seen this pattern before. In early technology shifts, it was easy to assume the hard problems were solved. Infrastructure was treated as finished, then usage surged. Systems built for a smaller world began to fail. Networks strained, power, phy
0:39 / 1:50
from transcript0:39 – 1:50
Have a podcast?
Run your podcast through Clypt
Get ranked clips, hooks, and ready-to-post copy from your own episodes. Free to try.