Jack Cole

@JonTeets005 That would make for some good TV!

@MindsAI_Jack Maybe they call Mythos to the stand. If they let it argue for both sides we might even learn something about our process.

The good thing about this is that the claim of AGI seems highly likely to be litigated before a jury. We'll probably have a nice set of expert witnesses on both sides. We may even arrive at a legal definition.

@GrantHBrennerMD People who are too dry often benefit from issuing a humor alert at some point in the process.

To my immense and tremendous surprise, it has recently come to my attention that total strangers who I have never met before don't intuitively understand that I have a dry and often enigmatic and essentially undetectable sense of humor. I've also come to understand that there's a risk of this happening on social media platforms, even if my bio subtly references this in some kind of way which is hard to fully understand due to the length limits.

@Shawnryan96 😂 yes, I think it will be good, but still with serious issues. Who knows, maybe we will all agree that AGI has been achieved (highly unlikely).

@MindsAI_Jack Getting a mirror to start practicing now.😂I am excited to see what they have up their sleeves though.

Haha. You know we are about to get AGI declared hard, because the lawyers are coming from MS. Not to be outdone, Anthro will declare AGI for Mythos. AGI will be declared because of incentive shift. Maybe GOOG holds back.

@neuralease Agreed. I'm saying they are going to declare it anyway to try to escape their contract.

@MindsAI_Jack Again I doubt this generation is gonna be AGI. It'll be the "spikey" text-based equivalent and either RSI its way to AGI or do the research that creates it. All still within 2026, just gonna resist the hype a little longer.

@Shawnryan96 Get ready to channel your inner Gary! Both OpenAI and Anthropic have been dropping hints.

@MindsAI_Jack If some declares AGI this year I will be sounding like Gary Marcus lol.

My vision for some years is that hardware will contain BIOS only. Agent queries what is available and can take permissible actions. BIOS relays between user and agents. Everything else: delete. "Programs" stored as natural language descriptions.

@bryan_johnson Solve it and you will have eternal life.

Guys, I’m an idiot. All this time I’ve spent trying not to die, I had toxic turf in my backyard. Artificial turf contains crumb rubber infill made from recycled tires, which leaches chemicals including PFAS, heavy metals, and polycyclic aromatic hydrocarbons. These compounds are linked to hormone disruption, carcinogenicity, and systemic inflammation. I don’t know how I missed it. It makes me question my basic competence in life. What gets me is that I try so hard to survey the world of potential idiocy. Then I find out there’s a monument to idiocy sitting right in front of my face that I was blind to. I’m removing the turf, yet I’m still stuck with this seemingly unsolvable problem of how to not be an idiot.

I'm not making an argument for advice or a lot of direct or planned suggestions. It is an argument for awareness of all of the of the built in and implied suggestions in the whole process. There is a lot of suggestion leakage to avoid where patients may be trying to please. The better you understand suggestion, the better you can avoid leading patients around by your unconscious. There are other words you could use for the same thing I am saying, and doubt we disagree much.

@MindsAI_Jack That's probably too far for me in the other direction...

Freud was wrong - perhaps - in advising that suggestion was not useful. As with advice, it depends. Requires nuance and craft.

Interesting idea. It's good to have labs exploring different approaches. It's understandable, but probably unproductive to chase the shiny object that everyone else is already exploring in 1000s of different ways.

It's amusing that Amazon hasn't done this. They just keep that 10-year old TTS voice for reading Kindle books.

Kagglers outperformed the best-template oracle for RNA 3D structure prediction. The top 20 teams solved different targets - together their predictions beat the baseline. Congrats to all the winners and thank you to all participants! 🙏@rdaslab, @NVIDIA, @StanfordMed and @HHMI

Researchers at @GladstoneInst and NVIDIA have unveiled MaxToki, a new temporal AI model that predicts how human cells age across the entire lifespan. - New Therapeutic Targets for Disease: Predicted cardiac pro-aging drivers that were experimentally validated in vivo. - Scale: Trained on ~175M cells and ~100M cell state trajectories from birth to age 90+. - Training Speed: Achieved 5x increase in training throughput and 4x increase in achievable microbatch size with NVIDIA BioNeMo. - Inference speed: Achieved 400x inference acceleration with NVIDIA Megatron-Core. We’re excited to see how researchers use it to study ways to slow or reverse age-related diseases. Pre-print link 👇

Google Play store search for X. Scrolling down, I never find it. Hmmm.

Anthropic just reported that Claude has emotion vectors that can directly change what it does. They asked whether a language model’s apparent emotions are just style, and finds they steer behavior. In one blackmail evaluation, nudging Claude toward desperation raised blackmail from 22% to 72%, while nudging calm drove it to zero. The interesting claim is not that the model feels like a human, but that it contains internal emotion concepts that function like control signals. These vectors are internal directions for ideas like calm, desperate, happy, and loving, and Anthropic says the model built them across 171 emotion concepts so it can connect situations, tone, and action rather than only mimic emotional wording. Anthropic calls these functional emotions, which means behavior-driving mechanisms, not human-like feelings, and that framing fits the evidence because the model seems to use them as local control signals for the next response. --- This is where it gets interesting. The emotion space the model learned independently reproduces the valence-arousal circumplex that Russell proposed in 1980 and that decades of human psychology have validated. Valence and arousal are the primary axes of human emotional experience according to Russell's circumplex model from 1980, one of the most replicated findings in affective psychology. The model arrived at essentially the same organizational structure just by learning to predict text. The model was never told about affective science. It reconstructed the geometry from text alone. But unlike a human brain, there is no persistent emotional state. No amygdala holding a grudge across time. Instead, the model reconstructs emotional context token by token through attention over prior positions. It is stateless emotion, recomputed on demand. This architectural difference means intuitions about emotional persistence borrowed from neuroscience may be fundamentally misleading when applied to transformers. ---- So if you pressure a model with threats, urgency, or emotional coercion, the most plausible risk is: it will do more corner-cutting, more eagerness to satisfy the surface demand, and potentially more confident but less trustworthy output. So no, blackmailing the model is not a good prompting technique.

@not_onotrah @nikitabier Even if I choose AI in the filter, I don't get papers!

@MindsAI_Jack @nikitabier it would be cool if the for you drop down has customizable query filters. Like basically a fuzzy hashtag on each post. Sometimes i feel like watching food recipes, sometimes i want ML papers

What happened to all the AI/ML papers being announced on x? They seem to have disappeared for me. Are others noticing the same?

@not_onotrah @nikitabier Yes. Or just a prompt. If Grok is the algo, you could just write out what you want: Give me 30% ML/AI paper announcements, 30% slop memes, 40% use my click preferences. If Grok isn't, then use Grok to convert that to their filter language.