Bring a normal robot near an MRI and this happens
That’s why the best way to see the brain and the most precise way to operate in it have never worked together
We just raised $8M @4DSurgical to change that
as far as I can tell, here are the stated views of tier 1 VC firms on extinction risk from AI so far:
- Denied/dismissed explicitly: @a16z, @khoslaventures, @benchmark
- Silent: @sequoia, @generalcatalyst, @ThriveCapital, @kleinerperkins, @IndexVentures, @Accel, @benchmark, @lightspeedvp
- Acknowledged the possibility: Thiel / @foundersfund, @GreylockVC
- Firm-level stance that avoiding human extinction is an active priority: none yet.
Who is going to lead from the front here?
We've all read the Hugging Face report, we are all staring down the cyber-offensive exponential, biological threats and attacks on physical infrastructure are not far off, and we are still early in navigating the threat landscape. There is no later time at which it will make more sense to act than today.
The frontier labs are already pulling out all the stops. Investors are a major power bloc at the frontier, they can see what's happening more clearly than almost anyone else. They need not see themselves as spectators to an unfolding drama, but instead can be -- and ought to be -- active participants in hardening civilizational defenses during the critical window (now).
That starts by taking ownership of the problem, in no unclear terms, and orienting towards sufficient (not cosmetic) solutions.
Abundant is partnering with @baseten to empower customers to own their capabilities.
Excited to leverage our experience working with every frontier lab to build environments tailored to the customers' unique needs.
99% of women and girls in Somalia have been mutilated in the following way: the clitoris is cut off, along with the inner labia. Then the inner flesh of the outer labia is removed before both sides are sewn tightly together, leaving a hole no bigger than a matchstick for urine and menses to trickle through.
All of this is done in childhood, and without any form of anesthesia whatsoever. Sometimes the little girl is given a scrap of leather or a stick to bite down on. The cutters use old razor blades, or sometimes a piece of broken glass.
This is only the beginning of a girl’s pain.
Most wedding nights in Somalia are aided by “the little knife” — that is, the new husband slices his bride open so that he can penetrate her now cut and bleeding scar tissue. Some men, however, simply batter their way through, tearing that tiny hole open by sheer force. When a woman goes into labor, she’s cut open more widely, then sewn back up afterwards so that when sex resumes she will have to be cut open yet again.
This practice leads to countless deaths from infection and a drastic increase in maternal mortality, but Somali men still refuse to marry uncut women, forcing mothers to have their daughters cut or face social stigma. Uncut girls are seen as dirty; the term for clitoris is a playground taunt in Somalia, and girls who are not cut are usually ostracized by their peers.
The Somali government recently debated the legality of this appalling custom, and the parliament largely split along gender lines. Female MPs roundly condemned FGM, while male MPs insisted that having a clitoris made a woman more difficult to penetrate, and quoted the hadith in which Muhammad condoned the practice. These are educated men who nevertheless claim that cutting a girl is necessary in order to control her sexuality, and that an uncut woman will be promiscuous and untrustworthy.
In spite of the Somali government’s continued support for FGM, foreign aid continues to flow into Somalia unobstructed. The West should refuse to give the Somali government so much as a single dollar until they outlaw FGM and take positive steps to combat the problem.
One tactic that has been highly successful involves showing Somali men a video of a girl being mutilated. Because FGM is seen as women’s business, most men have no idea what it entails. Men who watch this video are horrified, and the most common reaction has been “Not my daughter!” This proves that FGM is not hopelessly entrenched in Somali culture. Western pressure and resources could make an enormous difference in the lives of Somali women and girls, and put a stop to this barbaric practice.
Introducing Silico: the platform for building AI models with the precision of written software.
Silico lets researchers and engineers see inside their models, debug failures, and intentionally design them from the ground up.
Early access is open now. 🧵(1/10)
This morning, we are launching a seven figure ad campaign against @RoKhanna.
Today is the beginning of the end for @RoKhanna.
We are highlighting Ro's hypocrisy around $600,000,000 of trades since being elected to Congress.
You'll find it on Hulu, YouTube, Instagram, and eventually linear.
If you live in the district, you'll also be getting a mailer about the contrast of our policies.
We're in this to win it.
This is the most Orwellian thing I've ever heard.
@SFSuperiorCourt Judge Linda Colfax says "public safety is best served" by releasing a man who killed an Asian senior citizen as a joke.
@SFPDCommission says we can't see the killer's mugshot bc that would be "racist".
◉ We're live with 3 compute price tickers on 350,000 @Bloomberg terminals, and landed on the desk of every major CFO in the world.
On this note, it's time to bring @ComputeDesk out of stealth.
Compute is a financial asset class. We believe every company in the world will have an exposure to compute. So we are building capital markets infrastructure to price, hedge, finance, insure, arbitrage and market make compute.
In the last few months, we've built a mammoth compute price indexing engine, tracking bulk transacted compute prices, and the deepest compute price feeds in the market.
The tickers are the tip of the iceberg. Our market intelligence terminals are where GPU infrastructure operators and lenders go to price compute, expand margin and get ahead of their competitors. The biggest compute buyers use them to source capacity, and negotiate exclusive compute rates.
I would like to tell you our story and show you what else we've built. Something fun is coming.
@MongeMkt@Rainmaker1973 One group was people genetically predisposed to taking naps and the other was people who aren't. Since the groups aren't split by actual napping behavior, this concern shouldn't affect the results unless genetics meaningfully push people to take jobs that allow naps.
@hunterhhedge@ChronicBabak@semaforben@EmissaryOfNight Selective tanker passage fully benefits Iran and its allies, but has only a partial effect on stabilizing oil prices. It is significantly net beneficial for Iran.
.@GreatSkyAI just emerged from stealth w/ $14M Seed funding to provide >1,000,000x faster hardware for AI inference.
@JeffShainline & team spent the past decade at NIST re-designing AI hardware from first principles 🧵
https://t.co/ZErbBGxhLO
@EnergyAntonio@PauloMacro This sounds smart but breaks down when mapped onto the actual decision makers driving the war. Trump cares more about midterms and perceived economic health, and wouldn’t risk higher long term energy prices. Hard to imagine this being his strategy.
@goddreng10 Could be explained by marketing expansion to prime applicants who convert at lower rates. The absolute number of conversions is up, so they're not necessarily converting worse on their previous applicant pool
The intelligence law section of this is very persuasive if you don’t realize that every bad intelligence scandal in the last 30 years had a legal memo saying it complied with those authorities
a few scattered thoughts
(1) I think OpenAI has a good point that Anthropic's strategy of "ToS on edge-deployed guardrails-free models" is a weak safety device, and having an integrated safety stack is likely an improvement. It's still not clear to me what this could possibly mean in actual gray-area scenarios. Are safety teams building bulk “mass-surveillance” classifiers? Are FDEs going to sample traffic and hand-inspect prompts that look like surveillance, then confront the NSA about it? If they come across a batch of English-language phone or email intercepts, are they going to ask the NSA to prove the sender wasn't an American citizen? Some monitoring of American citizens is allowed — is OAI going to ask for details on the scope of NSA collection and make independent assessments of compliance with ODNI frameworks? Is OAI accounting for the fact that the NSA might operate adversarially to obscure surveillance that they think might run afoul of this contract?
(2) Just off the bat, it looks naive to have OAI stand there with the DoW and claim "no mass domestic surveillance since it's illegal" when we're sitting on 30 years of evidence of the NSA producing ad-hoc legal memos justifying extreme surveillance whenever it deems convenient. The DoW claims it doesn't do domestic surveillance despite being caught doing just that quite recently, and it appears that there's massive room for surveillance that practically violates Americans' assumptions of privacy even if it follows the letter of whatever ODNI framework is most recent. This seems to be the crux of Anthropic's objections — the NSA's definition of surveillance doesn't align with theirs or with the intuitions of your average citizen.
(3) The teeth of the agreement seem to be "if we find the DoW is in violation then we'll cancel the contract," but isn't the crux of the Anthropic saga that the DoW was upset that a deeply-integrated vendor resisted contract changes and pulled the nuclear option (supply chain risk designation) rather than let Anthropic walk away intact? To what extent can OpenAI actually enforce the guardrails here without running into the same existential scenario that Anthropic just ran into? It seems like a major point of contention between the DoW and Anthropic was Anthropic's assertion that they could pull the DoW's access if they deemed the ToS had been violated, which could leave the DoW out to dry in scenarios of critical national security... isn't that the power that OAI is reserving here as well? If the safety systems flag violations of the contract, they either need to stop the system from being abused (relatively quickly) to minimize harm, or they... just ask the DoW nicely to stop and then pursue a lengthy legal battle to get out of the contract? The contract either has teeth to prevent abuse, in which case you retain the ability to rug-pull the DoW, or it doesn't, in which case your safety systems are just there for letting you know how things are being used but don't have the power to actually stop abuse in any sort of reasonable time frame.
(4) Finally, and importantly: maybe none of my points above really matter? Sure, the NSA has a bad track record of abusing Americans' 4th amendment rights. Sadly, the DoW seems eager to produce whatever tenuous justification it needs in the moment to act on its whims (see double-tap strikes in the Caribbean, etc). At the end of the day, these are our collective problems to solve as a democratic society. The way to stop the NSA from another PRISM scandal probably isn't ToS on a SaaS contract, and it's probably not FDEs or overworked safety engineers sampling prompts from classified systems, it's by using our democratic system to elect officials who constrain these organizations from abusing their power according to the affordances outlined in the US Constitution. AI folks got their wish, monkey's paw curled, people outside the SF bubble woke up and realized the true power of general AI systems. There is no world where frontier AI labs refuse entirely to collaborate with the US Government. That's not going to happen — both because there's always a lab willing to lower the safety bar another foot (Grok's right there), and because we should want democracies to have the world's best AI. If you believe that the US and its allied democracies are humanity's best shot — and I do — then the goal has to be enabling them with the most powerful tools available while working tirelessly to constrain how those tools are used from within the system. Private enterprises were never going to be able to carte blanche dictate terms to the government. Yes, they have a decent amount of negotiating leverage now, but that leverage is declining and the USG always has the nationalization/DPA/etc nuclear options. If we want to stop the abuse of these extremely powerful tools by our governments, we need to do it through the channels afforded by our democracy, and that means crafting and enforcing laws that reasonably constrain how our government can use these models.
I'm not fully at peace with this last point. I’m worried about the authoritarian nonsense Trump is preparing to pull in the midterms. I’m worried about the fascist bent of DoW leadership. The 2020’s have not been a highlight of American democratic ideals, and we seem to be on a terrifying collision course between AGI’s arrival and some of the most contentious elections in American history. I still feel that at some level, you have to bet the farm on American and allied democracy prevailing. Part of that bet means enabling our defense complex with frontier AI, on terms that follow the letter of American law.
@bryan_johnson@grok evaluate the claim that dry sauna is superior to steam bath, referring to the arguments Bryan Johnson presents in this post. Speculate on whether steam bath can elicit similar a skin to core temperature gradient as dry sauna, and whether that even matters