> You likely replicated the sort of calculation they did. Demonstrated how to get oneself into trouble.
Yes, that was the entire point of my post, to give a specific example of what these LLMs are saying!
Not sure what you mean by that ?
I was just confirming the abysmal advice that AI, even something well regarded such as Sonnet 5, is dishing out.
Of course the teens would have been better off consulting a specialized hiking app and not AI, but there again adults have been following Google Maps and GPS directions and driving into lakes, so not so surprising.
The "add 30 mins" if you are doing it hopping one one leg is, well, interesting
Here is Sonnet 5's rather bad take on it:
> what would be a respectable time to walk 18 miles?
For an 18-mile walk, times generally break down like this:
Casual/relaxed pace (~20 min/mile, 3 mph): around 6 hours
Average brisk pace (~15â"17 min/mile, 3.5â"4 mph): around 4.5â"5 hours
Fast/fit pace (~13â"14 min/mile, 4.3â"4.5 mph): around 4â"4.2 hours
Competitive race-walking pace: could dip toward 3â"3.5 hours, but that's serious effort
For most reasonably fit people doing this as a single long walk (not a race), finishing in around 5 hours, including a few short breaks, would be a solid, respectable result. If you're doing it on hilly terrain, in heat, or with a loaded pack, add 30â"60+ minutes.
***
So
Doing it on hilly terrain, with a loaded pack, on a hot day
Hmmm
FWIW, he (and his fund) haven't lost everything. After this recent rout, his fund is still up 80% for far this year (before this huge drop it had been up 400%).
No doubt his risk management is going to be better going forwards. It'll be interesting to see how he does going forwards - his bets to date seem to have consisted of going all-in on everything related to datacenter growth, which was pretty obvious at least as a short term bet. Let's see if he can predict the effect of AI on the economy and continue to do well.
True, but there were (and still are) at least two reasons for investors to stick with SALP and Aschenbrenner.
1) He is extremely well connected in SV. He's ex-OpenAI, and shares a house with Sholto Douglas (high level Anthropic employee), Dylan Palel (of Semi Analysis - who are essentially doing the research that SALP, and others, are investing on), and Dwarkesh Patel.
2) Despite this setback, he has still made tons of money for his investors, even this year, and may well continue to do so. Many high profile investors such as George Soros have also made large strategic bets that lost tons of money, but still recovered from it. Aschenbrenner has no doubt learned a painful lesson in risk management and use of leverage.
Even after this 67% drop, SALP is still up 80% so far this year, having been up over 400% earlier. I'm sure many investors are sticking with him.
How about rather than trying to be a smart ass, or the president trying to blame a Democratic governor for an Iranian state-sponsored attack, the president actually does his job?
Is that too much to ask?
Ask like a president - be a leader, not a whiner. Take charge. Tell your cyber-Tsar to have a meeting with the state governors, or appropriate representatives, and get appropriate protection measures in place for all critical infrastructure (not just water treatment plants).
Yeah, it's not rocket science, unless you are a complete fucking idiot.
But it doesn't line up.
There is an old paper (by Bernard Barrs) "In the Theatre of Consciousness" that likens the mind full of ideas to a theater's stage, with consciousness being the spotlight that illuminates some ideas, leaving others dark. This theory of consciousness refers to itself as "Global Workspace Theory".
All Anthropic have done here is choose to call the middle layers of a Transformer a "Global Workspace" (may as well have called it a "Pink Wombat") and say "wow, look, we've got a workspace too, so it may be conscious".
The spotlight of Barr's theory is the bit corresponding to consciousness, so does a Transformer have anything that can be called a "Spotlight" to try to make some analogy here? No.
Necessarily a transformer has many language patterns/predictions internally that aren't in the output it generates, since it is trained to predict all possible continuations of the input and therefore predicts a bunch of alternate next tokens, from which the framework calling the model picks one AT RANDOM.
So Baar's workspace/theatre of ideas has consciousness shining a spotlight on one idea, and a Transformer has a freakin' random number generator picking one.
Do you consider that "lining up"?
It'd be like saying that a banana and a car are similar because you can call a banana skin a "car body", and ignore the fact that the banana has no engine and can't move.
Nah - completely wrong.
Find a model that does this, and ask it to SPELL strawberry instead, and I'll guarantee it will get it right.
The problems some models have (esp. if you don't allow them to think out loud) is COUNTING, not "mapping tokens to letters".
The reason they don't have a problem doing this mapping is because to them it's not mapping - it's prediction, which is the one thing they have been trained to do. They are predicting that the input "how do you spell strawberry?" will be followed by 'S', 'T'
> neural networks are definitely doing something
No shit, Sherlock !
And what language models have been trained to do is find language patterns that help predict next token.
Anthropic : Holy fuck! This thing is finding language patterns that help predict next token! It's alive! It's worth a bajillion dollars!
Yes, although Baars' "Global Workspace Theory" (pertaining to the human brain and consciousness) is really irrelevant here. Anthropic only refer to it because they want, as usual, to make grandiose claims and want you to believe that LLMs may be conscious (while of course not being willing to commit to an actual definition of consciousness).
Of course Baar's theater (not ocean) metapor is central to his theory of consciousness which he compares to a spotlight in a theater drawing attention to some players on stage while ignoring others.
OK, so if you strip away Anthropic's poorly applied comparison to Baars' GWT, and appeal to you to believe that LLMs may be conscious, then what are we left with? What is the actual technical content of their blog post (which the Slashdot story doesn't even bother to link)?
https://transformer-circuits.p...
Their title "Verbalizable Representations Form a Global Workspace in Language Models" cuts to the heart of it.
The "Workspace" that Anthropic are referring to (fig 2) is just the the activations of the middle layers of a Transformer which is where most of the high level pattern recognition and prediction occurs (since the input/output layers need to convert word sequences to/from these richer middle layer embeddings).
OK, so why call it a GLOBAL workspace (which is central to their grandiose brain analogy)? A: Because the diagram of a Transformer you are likely familiar with puts all the focus on the stacked Transformer blocks with their attention and feed-forward components, and presents the residual connections as more of a technicality (maybe you think of these as just to help gradiant
propagation as in a ResNet). In fact the residual connections are really central to a Transformer, which may be better thought of as an embedding highway (the "residual" connections) passing thought the layers, with the job of each layer being to incrementally augment these embeddings by adding data to them (derived from attention and feed-forward blocks).
So, what makes these middle layer "workspace" activations (i.e. embeddings) global is that the residual highway interconnects all layers and anything added in lower layers will therefore be globally accessible to all layers above it.
So, there you have it: Surprise surprise (NOT) Transformers share embedding values across layers (woo hoo - global) which represent high level concepts, some of which (depending on random output token selection) may manifest in the output, and others just representing internal abstractions and output paths not taken.
Anthropics PR spin: brainz.. brains.. it's alive! it's conscious!
I guess you could frame it like that, but knowing how it works doesn't help if you don't have the know-how to build it.
How do you increase your EUV power from 100W to 1000W (this took ASML years to figure out)?
How do you make mirrors smooth to within a single atom deviation?
How do you make chemical etch pure to the parts-per-trillion level?
No doubt the Chinese will figure these things out by themselves if they have to, but I'm sure there are a few secrets to be had that would speed that up!
EUV would be nice to have, but at the end of the day it all comes down to cost. Without EUV your chips will be larger and slower, so you'll need more of them (more expensive) to build a cluster of the same power.
The cost of serving (not price to you) an AI model is mostly hardware depreciation cost, not operating cost, with the accelerator chip cost being a large part of that.
NVIDIA's H100 chip costs $25-40K
Huawei's comparable Ascend 950PR costs $7-16K
As you can see, lack of EUV is not stopping the Chinese from being price competitive.
That's irrelevant. Sure China has been blocked from buying ASML EUV machines, but they are doing just fine with DUV and companies like SMIC (Chinese semicondustor fab cf TSMC) have pushed it to ~7nm node size.
The Chinese are making their own AI accelerators, such as Huawei's Ascend series, which DeepSeek are using, but just like OpenAI making their own chips to avoid the NVIDIA tax, DeepSeek now want to make their own presumably at least in part to avoid the Huawei tax (as well as perhaps to gain even greater efficiency).
This isn't intended for consumer use at all - they are just doing the same as all the AI companies and trying to lower their inference cost by not paying a profit margin to someone else (in this case Huawei).
"The medium is the massage." -- Crazy Nigel
