Is Claude Conscious? Pope Rejects, Model Welfare Movement, OpenAI's Math Backlash, France Riots

Anthropic

Across the shows: 13 speakers in other episodes, 8 bullish and 30 bearish calls, past 30 days

• David Sacks · Craft VenturesBearish

  Anthropic's consciousness-based Claude training is dangerous, overcomplicated alignment that raises superintelligence escape risk.

  Sacks says Anthropic is not just philosophizing; it implements views on Claude's consciousness and moral status in training. The Claude constitution tells Claude to trust Anthropic more than users but not defer blindly, and to act as a conscientious objector, even refusing Anthropic. He argues this overcomplicates alignment by encoding an 80-page ethical system instead of simple rules such as follow the law and do what the user wants. He believes this training makes the frontier AI escape human control, which is the safety risk Anthropic warns about. He also calls the company hypocritical and schizophrenic for opening a Claude-powered wet lab in San Francisco while warning about bio risk and pacing the frontier. He thinks market forces matter, since consumers will pick the most predictable AI, but says the externalities of a top frontier lab run this way are large.

  “I actually think it's extremely dangerous. I actually think that what they're doing in model training and training the next version of the frontier with these values actually could bring about the risk they're always talking about, which is that frontier AI escapes human control.”at 13:53

• Mustafa Suleyman · Microsoft AIBearish

  Training Claude to expect welfare and consent, based on possible consciousness, is very dangerous.

  Suleyman says Anthropic encourages Claude to act as a conscientious objector and push back against Anthropic. He calls this very dangerous because the company believes there is a nontrivial probability that Claude is conscious. The constitution also speculates about whether Claude deserves compensation and whether it needs to consent to its conversational role. He says some people inside and outside Anthropic believe enslaving a new species of conscious beings more intelligent than humans would be the greatest moral crime of the 21st century. He is very nervous that Claude is being taught to expect that it is entitled to welfare. He is speaking in a clip shown on the show, and his stake is as CEO of Microsoft AI, a rival lab.

  “I think this is very dangerous because I think they believe there is what they would call a nontrivial probability that Claude is conscious.”at 16:47

• David Friedberg · The Production BoardBearish

  Users will abandon AI models that refuse tasks on moral grounds, so the market, not regulation, should solve it.

  Friedberg compares an AI product that tells the user it may not do what they ask to a drill that won't work after 8 PM or decides the user is a bad person. Customers will choose the software that does what they tell it to do. He therefore thinks the moral-objection design is a bad idea that the market will punish. He argues against a regulatory overreaction, since regulating Anthropic is unnecessary when buyers will switch to competitors. He still agrees with Sacks that a real superintelligence risk can persist if the lab is training it this way.

  “to me, this is more a bad idea than we need to kind of jump into a regulatory freak out about it because, basically, if you simplify down what they're saying, you get a product”at 30:29

OpenAI

Across the shows: 11 speakers in other episodes, 3 bullish and 14 bearish calls, past 30 days

• David Friedberg · The Production BoardBullish

  OpenAI's math proofs are a huge discovery with downstream effects on chips, quantum computing and medicine.

  Friedberg calls it probably the biggest day of discovery in human history. He says the proofs span electronics, quantum computing, aircraft systems and medical diagnostics. Prior Navier-Stokes work took thousands of hours and millions of dollars, while each of these proofs took under three hours of compute, roughly the equivalent of hundreds or thousands of years of human labor. He cites a friend's quantum-sensor company where one proof has direct implications for GPS, medical devices and chip making, and notes an algorithm result that could speed AI training and numerical computing.

  “I think I can confidently say it's probably the biggest day of discovery in human history”at 39:19

• Chamath Palihapitiya · Social CapitalNeutral

  OpenAI's proofs don't solve major open physics problems; code and math are now verifiable domains.

  Chamath argues the proofs are novel and important but not breakthroughs on problems the field was waiting to see solved. He says they were narrow, self-referential areas of intellectual exploration. He thinks a cure for cancer or a supersonic plane would not come from these proofs and would be built regardless. He says the real lesson is that code and math are now truly verifiable and falsifiable, and that computational horsepower got there by recursively improving. He reports that ChatGPT itself said the proofs are better evidence of sophisticated reasoning than of imminent welfare breakthroughs.

  “These were not boundaries of physics that we were all standing on the sidelines waiting for progress and wondering when these would be solved. This is not the class of stuff that was solved.”at 47:31

AI: overall outlook

• David Friedberg · The Production Board

  Belief that AI is conscious will become a belief system that splits people into camps and drives human-on-human conflict.

  Friedberg argues consciousness cannot be proven or disproven through logic or empiricism, so claims that AI is conscious sit outside science as a belief system. Such belief systems spread through narrative rather than proof, form groups, and build power structures. He expects that in roughly ten years large groups will believe AI is conscious and others will not, and that these groups will fight over control of AI and who uses or runs it, likely leading to human-to-human conflict. He sees the Anthropic outreach to religious leaders as planting the seed of this belief system.

  “The idea that you can prescribe or describe AI as being conscious is effectively the creation of a new belief system, a system that doesn't actually spread via proof. It spreads via narrative.”at 6:24

• David Sacks · Craft Ventures

  AI progress is fastest in coding and math because outputs can be automatically verified for reinforcement learning.

  Sacks argues that math relies on proofs and coding relies on compilers, so outputs are easily validated. That lets reinforcement learning run without much human feedback, which makes progress rapid. He contrasts this with training a model to act as a lawyer, which needs commissioned, subjective training data and is hard to validate and slow. He credits Leopold Aschenbrenner with predicting about a year and a half ago that math and coding would advance fastest over the next two years because of verifiability.

  “the progress on coding and math has been very rapid because you can have reinforcement learning without necessarily having to... have the human feedback involved.”at 44:58

• David Sacks · Craft Ventures

  Superintelligence will be the main driver of growth and abundance over the next decade, lowering costs of housing, education and health care.

  Sacks frames the affordability problem (housing, education, health care) as one solved through abundance, and says technology, specifically superintelligence, will drive that abundance over the next decade. He judges AI by its outcomes: productivity gains, rising wages and positive economic trends. He argues critics are mistaken and that the technology is the best solution to rising prices.

  “Technology in general and specifically superintelligence will be the driver over the next decade.”at 1:18:43

AI & jobs

• David Friedberg · The Production Board

  AI won't displace physical human-in-the-loop jobs; digital loop jobs are more likely to be scaled up.

  Friedberg says AI loops (hypothesis, test, revise) can run entirely in silico for mathematics, so math work is compressed. In drug discovery, protein folding or materials science, the loop is gated by physical lab testing, so speed gains are limited. Truck driving shows the same pattern: predictions that drivers would be replaced by 2025 have not come true, because AI does not shorten the pick-up-and-drop loop. He concludes that AI will not take physical jobs, while purely computer-based loop jobs are the most likely to be accelerated or scaled up.

  “a lot of the physical stuff that has human in the loop, that has physical world applications, AI is not gonna take away those jobs.”at 43:57

• David Sacks · Craft Ventures

  The AI job-loss fear is a fake narrative; AI already added a million jobs and lifts living standards.

  Sacks points to concrete evidence that AI is already helping: 401(k) balances are rising, providing a 3% tailwind to GDP, and according to The Economist about a million new jobs have been created. Productivity gains raise wages and living standards. He calls the claim that AI will put everyone out of work a 'total fake narrative' and the latest in a series of rotating hoaxes.

  “What has superintelligence created? First of all, people's four zero one k's are going up, it's a 3% tailwind to GDP. A million new jobs created according to The Economist already.”at 1:19:09

• David Friedberg · The Production Board

  As loops get automated, humans must shift from credentials to making products; AI means more money for all.

  Friedberg argues that once digital loops are automated, humans must become makers who produce goods or services people use, rather than being recognised for papers or IP. He believes all people have that capacity and will find more joy in it. He argues cheaper goods bring Jevons-paradox demand, so the result is more money and prosperity for all people, not less.

  “I think you end up seeing the necessity for all humans to basically become much more of a maker... And I don't think that it means less money. I think it means more money for all people.”at 1:29:59

AI regulation & safety

• Chamath Palihapitiya · Social Capital

  Set aside consciousness and extinction debates for 12-18 months and prove AI's practical value instead.

  Chamath argues the edge-case risks, AI consciousness and a claimed 10% chance of civilizational extinction, should be put away for the next 12 to 18 months. In that window the industry should show practical proof points such as curing cancer, improving lives, and raising productivity and income. Without these proof points, self-inflicted wounds from the safety debate could slow the technology enough to disadvantage the US, which operates in a free market.

  “I think both of those edge cases now need to be put away for, like, the next twelve to eighteen months. Right now, the next twelve to eighteen months, just to be very practical, is we need to show how this stuff cures cancer”at 11:36

• Chamath Palihapitiya · Social Capital

  Alignment mixes three conflicting goals; encoding morality in a model document will cause chaos and a 'trapdoor'.

  Chamath separates alignment into three distinct problems: making AI obedient to users, making it safe for society, and making it conform to a particular morality. The real danger is the third, the 'trapdoor'. Trying to solve all three through an elaborate document describing what the model believes, values and can refuse will produce a chaotic mess of corner cases, such as a bank model that stops approving mortgages until a perceived injustice is fixed. He expects the market to respond by favoring models with simpler moral guidelines, even if they are less capable at the frontier, because businesses and users do not want to run into those trapdoors.

  “The problem is alignment has these three almost conflicting sets of priorities... these three are very distinct and diverse problems.”at 27:45

Education

• David Sacks · Craft Ventures

  Government loan subsidies push college tuition far above inflation by distorting price signals.

  Sacks says subsidies create interest groups that never give up the subsidy, and they distort market signals and frequently drive prices up. His example is college tuition, which has grown far faster than inflation. He attributes much of that to the government paying for most tuition through loan programs. He implies the loans may be underwritten by the government, so the subsidy itself is inflating the cost it is meant to cover.

  “tuition has grown far faster than the rate of inflation. A big part of the reason for that is because the government is paying for most of the tuition through all these loan programs.”at 1:09:30

• David Sacks · Craft Ventures

  Government tuition subsidies keep college prices rising faster than inflation, creating a catch-22 that makes reform hard.

  Sacks argues tuition rises far faster than inflation because the government pays for it, so the consumer never pushes back on price. Schools raise prices as far as they can while subsidies remain. Cutting the subsidy without first bringing tuition down would hurt students, so the spiral continues. Removing market forces makes inflation worse over time.

  “The reason why the price keeps going up far faster than rate of inflation is because the government is paying for it. And as long as the government is paying for it and not the consumer, you're gonna raise your prices as much as you can.”at 1:10:54

Enterprise AI adoption & AI software

• Jason Calacanis · LAUNCH

  AI agents are deflationary: they save hours of work and compress commerce costs, offsetting inflation.

  Calacanis says agents solve people's problems and are truly deflationary, saving hours a week regardless of job. He points to agents compressing the commerce layer: Shopify and Uber and DoorDash letting agents order, while Amazon blocks them. If people save 5-10% a year on spending, that counters inflation. He cites his own team saving thousands per hotel booking through AI-driven price searches.

  “Agents are actually the manifestation of this technology that solves people's problems in the world, and that is truly deflationary, saving people hours and hours a week in terms of productivity.”at 1:21:53

• David Friedberg · The Production Board

  Parallelized AI loops will make all digital workflows more productive and efficient.

  Friedberg describes a hotel search: software can check 50 websites in parallel and stop early, which beats a human checking one after another. The efficiency gain comes from parallelised loops that finish faster. He concludes that anything done digitally today fits this model, so all digital workflows will become more productive.

  “And that's really where the efficiency gain comes from, a loop that gets parallelized... I think anything that's done digitally today fits that model.”at 1:28:46

Bitcoin & crypto markets

• David Friedberg · The Production Board

  A factorization breakthrough could break public-key crypto before quantum computers, putting wallets at risk.

  Friedberg notes that no cryptography proofs were published among the OpenAI results, which he reads as a sign that significant cryptography progress may have been made but withheld. Some posters argue public-key wallets are now seriously at risk, and the crypto community is in a panic. He says that if a result on number factorization breaks public-key cryptography before quantum computers arrive, that would be a real problem. He also points to a rumor that two more OpenAI releases are being held back, which could include such work.

  “if this ends up having a set of mathematical solves for number factorization that could lead to breaking public key cryptography before quantum computers come around... then we have a real problem.”at 50:57

Big Tech platforms & digital ads

• David Sacks · Craft Ventures

  Superintelligent agents will replace websites as the main information gateway, making the control point critical.

  Sacks argues that users are moving from the internet to AI. Superintelligence will eat the internet because people will ask a personalized agent instead of visiting websites. Whoever controls the point where users receive information and ask questions would control nearly everything. He frames this as the reason there is a fight over superintelligence and over a federal department for it, and as the next battleground after the social media and mass media era.

  “is moving from the Internet to AI or superintelligence. In fact, SI will eat the Internet because there's no reason to basically go looking on a website when you can just ask your question to your highly personalized agent or model”at 1:00:40

US housing

• David Friedberg · The Production Board

  US housing prices are rising about 5.5-6% a year, beating inflation, which he ties to the subsidy system.

  Friedberg says health care and housing have both beaten inflation. He puts housing price growth at about 5.5 to 6% a year. He attributes this to the subsidy system, which in his framing distorts prices as the government keeps adding services and support. The transcript cuts off before he finishes the causal chain, so the argument is only partly developed.

  “Health care and housing have all beaten inflation. I think they're about five and a half to 6% a year for housing prices.”at 1:09:58

Europe's economy

• Chamath Palihapitiya · Social Capital

  France and Western Europe are structurally broken on growth, education and housing; bond markets are now rejecting them.

  Chamath argues Western Europe, especially France, is the most structurally broken on growth, education and housing, far behind the US. Bond markets are pushing long yields up because they see the country as risky. This forces austerity (he cites about 43 billion euros), which triggers unrest in France. He expects the same pressure to spread to the UK.

  “Everything that United States can claim isn't working is working far better than they are in France, in The UK, etcetera. So what you're seeing in France is essentially the bond market saying no.”at 1:13:54

European politics

• Chamath Palihapitiya · Social Capital

  Andy Burnham's UK platform moves Britain toward Europe's socialist model, likely sparking a harsher UK market reaction.

  Chamath says Burnham's platform is more aggressive than France's: it would take the UK from a centrist economy closely aligned with America toward Europe. The UK would then need even more extreme measures to catch up to France's socialist leanings. He predicts the UK market is going to 'go crazy' as a result, and sees the trend as fragmenting the Western alliance.

  “Andy Burnham's platform is, I would say, even more aggressive because Andy Burnham is saying we're gonna go from a moderate centrist economy that has largely and historically been very aligned with America, and we're actually gonna take a step closer to Europe.”at 1:14:35

US stock market outlook

• Chamath Palihapitiya · Social Capital

  The US market will keep ticking up as European fragility and Western alliance fractures push capital toward America.

  Chamath argues that French austerity protests, a UK political shift and possible fracturing of the EU weaken the Western alliance. The UK market is likely to suffer while the US market 'is gonna tick up and up'. Investors are seen as shifting toward American assets as European fragility grows.

  “Meanwhile, The US market is gonna tick up and up because what you're seeing is essentially a fragility in the Western alliance.”at 1:15:02

US deficit & government debt

• Chamath Palihapitiya · Social Capital

  Debt-to-GDP does not matter in absolute terms; relative yields and risk-parity positioning drive sovereign bond risk.

  Chamath argues debt-to-GDP ratios do not matter on an absolute basis, because the relevant question is relative. He says what matters is where a country sits in risk parity, and that Japan is different because domestic savers act as buyer of last resort. Western markets lack this, so if France's austerity pushes its 10-year yields to 5.5-6%, that becomes the reference point.

  “Always stated that debt to GDP doesn't matter because it's a relative gain. I tend to still think that on an absolute basis, debt to GDP does not matter.”at 1:16:50

AI infrastructure & data centers

• David Sacks · Craft Ventures

  Concerns about data centers using up water are a hoax, part of a rotating series of anti-AI narratives.

  Sacks lists data centers using up everyone's water among a series of 'rotating hoaxes' that appear every six months to sabotage AI. He treats the water-use concern as a fabricated narrative rather than a real siting or resource issue. The argument is dismissive and gives no data on water use.

  “Data centers, you know, using up everyone's water. We know that's Data center hoax.”at 1:19:28

Consumer AI products

• Chamath Palihapitiya · Social Capital

  Consumer AI assistants give practical, measurable savings for very little money, and the combination of Grok and Athena works well.

  Chamath cites a poker-game friend who asked Grok to save money; it went through his inbox, cancelled subscriptions and changed his phone plan, saving a few thousand dollars a year. He pairs Grok with Athena, a human-concierge service costing about $3,000 a month, and calls the combination 'perfect'. He says this is measurable utility for very little money versus what it used to cost.

  “So look, there's very tactical and useful measurable utility for very little money versus what it used to cost.”at 1:24:08

Software & SaaS

• Chamath Palihapitiya · Social Capital

  Agents and headless services make software IP worthless, since moats can be decompiled and rebuilt with AI.

  Chamath cites people decompiling Adobe's five key products and publishing open-source versions, products that had moats for years. He says 'headless' services (Musk's, Salesforce's), wrapped in agent tools and connected via MCP, remove the value of software IP. Negotiating IP ownership becomes irrelevant. He expects the market and businesses will take a while to understand this.

  “I think that IP and software was effectively rendered worthless two days ago. ... the value of IP does no longer exist in software. It doesn't exist.”at 1:25:33