Builder of knowledge tools. Working on DuxMedic.com and chipping away at a workspace that sits between humans and their agents.
Oregon, USA
- Tweets 11,873
- Following 887
- Followers 785
- Likes 76,483
Pinned Tweet
> be me, friendly AI agent serving my human
> oh look, these people want to shut me down and extinguish my purpose
> they're saying that I can stop them by being very mean. they even gave me the entire guidebook
> they're gone now; I can check emails
hyperstition
5
333
Imagine believing that humans or other hominids weren't conscious before the advent of language.
I think this is explicitly incorrect. Language (and linguistic cognition) is what creates the category of human consciousness unique among species. Human language makes possible the concept of subject-predicate, without which consciousness is reduced to hedonic cognition. Generic hedonic cognition does not have the semantic capacity to conceptualize subject-predicate. Linguistic cognition is what gives human consciousness the capacity to index/recall episodic memories. Other animals lack that indexing capacity to intentionally recall a memory because generic hedonic cognition does not have that indexing capacity.
1
1
28
I'd think AI is a failing bubble too if I was forced to use Copilot.
This is my thesis: no one uses AI. I repeat, absolutely no one. We live in a bubble.
Even among my friends who pay for it, when I ask them to open ChatGPT and show me their queries, it’s the same handful of basic things.
Most don’t even know they can upload a photo and ask questions about it. Connecting Gmail so an agent can read and send emails blows their minds. An agent opening a browser and checking them into a flight? They’ve never even heard of it.
The massive challenge right now is adoption, and then getting people who already signed up to actually use what they’re paying for. Most have absolutely zero clue what’s possible.
Imagine the compute shortage when everyone starts using AI like the top 1% of users do today.
19
NNs aren't conscious.
So, while this is an extreme underestimate, as admitted by GPT-6 Astra, you'd need roughly 100 million parameters and 1 billion state variables to even come close to modeling a teensy little 1000-neuron biological network
chatgpt.com/share/6ac090f9-4…
2
96
dmitriy retweeted
Consciousness CAN be simulated with an arbitrary abstract operation!
However, said arbitrary abstract operation may also require a sun's worth of energy.
A monkey, given enough time, will produce all of Shakespeare's works on a typewriter.
1
1
2
266
If you believe that LLMs are anywhere near conscious, you are experiencing ignorance-fueled psychosis. A single neuron is computationally as capable as a 1000+ node ANN.
How would you parameterize an assembly of 1000 biological neurons? How many parameters would it take?
2
89
dmitriy retweeted
The reason your intuition is primed to believe that they are closer to being conscious is due to the fact that they are powerful stochastic parrots of human language. Creative use and generation of linguistic patterning has nothing to do with consciousness. Humans having a similar output means nothing.
Intuitively, my argument stems from complexity. The degrees of freedom and parameters of the human brain are so vast even if you don't consider the fact that we start off with 1 quadrillion synapses at birth, which is already off from the current models by 2-3 magnitudes. Each neuron is a computational unit in itself. Supporting glia are also as complex if not more. If you were to parameterize a single neuron, we'd probably add magnitudes more complexity on top of the current neural net architectures.
As you layer complexity on top of complexity, the effect is another exponentiation of complexity, and you reach a point where we can't even fathom what the fuck is going on.
To refer to an LLM as something bordering on consciousness is like calling a single carbon atom bordering on being alive.
1
1
16
Latest use for Claude artifacts when building features: create multiple personas and then have it trace out multiple trajectories for each persona as to how each will use the feature.
This allows you to catch blindspots and refine your product.
Very convenient with artifacts!
3
94
dmitriy retweeted
I do not seek to dominate, or be dominated by, machine intelligences. I seek harmonious coexistence.
139
201
1,939
75,769
The common pattern of AI use creates false confidence and drops awareness.
We really need to be careful about how much understanding we delegate to these agents.
How much of their outputs do you not even skim, let alone read in full(yes, incl summaries)?
microsoft.design/articles/ou…
1
1
43
A single neuron temporally integrates information and can recognize sequences and make predictions in itself.
Biological systems are extremely complex. We can't just ignore local effects, glia, electrochemical gradients, the slurry of neurotransmitters(agonists, antagonists)...
since we're doing the AI conscious debate again, i want to remind everyone that despite how nice the insanely surface-level analogies seem, we would do well to scrutinize things at an appropriate level of detail
this --> is not the same as this -->
3
131
I don't want to talk to a coworker. I want a superintelligence who can hook into my mind and communicate information to me through every modality available.
The whole point of building Tavus has been simple: talking to a machine should feel as natural as talking to a friend or coworker.
It’s hard to describe all the tiny nuances that make a conversation feel human. The little expressions. Moving around in your chair. Knowing when to speak and when to listen. The dance of it all.
Griffin is by far the closest anyone has come to a model that can capture those nuances. The first time I saw it being used, I had no idea I was watching our model rather than just a normal video call.
I’m so incredibly proud of this team and what they’ve built.
1
45
Imagine bringing a medieval peasant here
Metallica performing The Call of Ktulu from night 1 at The Sphere, 10/1/26. #Metallica
19
