mckirk

16 ms·

181.
by mckirk 3y ago
The problem is that these models do not have any working memory they could use to carry out such tasks, which are on a meta-level when seen from a language perspective. They can only go with their 'gut instinct' for selecting the
182.
by mckirk 3y ago
True, I should probably have been more specific, with 'LLMs' I was just referring to the current Transformer based models. And yes, we could of course get the right answer to PARITY problems, but only if we were allowed some kind
183.
by mckirk 3y ago
The thought I had while reading, which you also touched on towards the end, was: You might indeed not need a model powerful enough to compute PARITY in 'one shot', if what you are aiming for is 'only' modeling human thou
184.
by mckirk 4y ago
I know, it's awfully handwavy, but it's at least one possibility that would 'make sense', and I haven't really found any others, though I've thought about it quite a bit. Interesting side note though: If you ac
185.
by mckirk 4y ago
Humans didn't have any incentive to develop 'actual' consciousness either. Evolution simply incentivized 'better adaptation to the challenges of the environment', which produced consciousness as a side-effect. But f
186.
by mckirk 4y ago
The problem is that memes, at least the ones that really take off, work because they are creative and funny, and so far that is probably the area AIs like GPT4 struggle the most with. It makes sense when you think about it, a technology tha
187.
by mckirk 4y ago
Also marzipan; was told by somebody working at an airport that if you want to look _really_ suspicious, just bring a block of marzipan right next to your headphones.
188.
by mckirk 4y ago
SuperStonk is still around? Have they pivoted from 'GME is gonna take off just wait for it any day now', or is that still the gospel?
189.
by mckirk 4y ago
We'll have to see whether there even still _is_ a tick rate. Afaik you only need ticks if you need to simulate something in your world that needs iterative updates, such as objects following a complex motion that doesn't have clos
190.
by mckirk 4y ago
I'd say it's a lot closer to Valorant than to the max-saturation graphics of Fortnite. But I think they found a good balance, it still looks like you can take the world seriously imho.
191.
by mckirk 4y ago
Amazing to finally see some work being put into CS, and just this smoke grenade sneak peek already makes me very curious to see how this will play in practice. It's clear at least that they aren't afraid of changing the meta in a
192.
by mckirk 4y ago
I can confirm that with the improved auto-pilot it is possible to get a 103.6 point landing after performing 1,151 flips (and reaching a max speed of 504 mph and a height of about 38k feet), if you can spare your phone for 38 minutes.
193.
by mckirk 4y ago
That highly depends on your frame of reference, I suppose. In most of Germany that would be considered fairly 'heavy' use already, whereas most of the West Coast would probably consider that to be rookie numbers.
194.
by mckirk 4y ago
What kind of evidence would you need to convince yourself of an AI being sentient?
195.
by mckirk 4y ago
"AI therapist" certainly seems like a new job that will see plenty of demand in the future, if Bing Chat is any indication.
196.
by mckirk 4y ago
Let's just say 'it will have been dragged' to be on the safe side either way.
197.
by mckirk 4y ago
I am saying that GP's claim of "just because it only uses some basic math, it cannot be sentient/conscious/intelligent" does not seem valid to me. The only way such a claim could be made, as far as I can see, would
198.
by mckirk 4y ago
I concur, which is why I find it funny that you brought religion into this in the first place. By asserting that "people entertaining the thought that this neural network could be sentient/conscious must be misguided, and the only
199.
by mckirk 4y ago
Where's your proof that your own brain is doing anything that couldn't be represented with a 'matrix multiplication'?
200.
by mckirk 4y ago
There's two kinds of 'memory' in these models, the short-term memory that consists of the tokens of the current conversation, and the long-term memory that's encoded (in latent space) in the neural network's structu
201.
by mckirk 4y ago
There's always one neutron star with no obvious place to go left in the end.
202.
by mckirk 4y ago
> So a dumb human is not sentient? /s That was just my way of preempting any 'lol of course ChatGPT isn't sentient, look at the crap it produces' comments, of which there thankfully were none. > But we should under
203.
by mckirk 4y ago
You have been a good user.
204.
by mckirk 4y ago
That is what I'd call empathizing though. You can 'put yourself in the other person's shoes', because of the expectation that your experiences are somewhat similar (thanks to similarly capable brains). But we have no ide
205.
by mckirk 4y ago
I've thought about this as well. If something seems 'sentient' from the outside for all intents and purposes, there's nothing that would really differentiate it from actual sentience, as far as we can tell. As an example
206.
by mckirk 4y ago
If you've ever looked at the top submissions for any coding competition, you know that the premise of this study is flawed. I assume that the people who score well on these competitions are quite capable and would write good code in ot
207.
by mckirk 4y ago
Great book! I can also wholly recommend the audiobook, it was quite well done.
208.
by mckirk 4y ago
Presumably neom's initials (check his profile), which would be shown beside the prompt.
209.
by mckirk 4y ago
We lüft the house in the morning, we lüft the house at night. We lüft the house in the afternoon, it makes us feel alright. We lüft the house in times of peace, and we lüft in times of war. We lüft the house before we lüft the house, and th
210.
by mckirk 4y ago
Well, now that link is dead as well apparently, so I'm starting to believe there is indeed some info here that wasn't supposed to be public (yet).

More ›