Jul 13, 2026 · 2h 2m
OpenAI whistleblower warns superintelligence could arrive by 2029
OpenAI Whistleblower FINALLY Speaks: “AI Has A 70% Chance Of Going Horribly Wrong!“
As tech giants race to automate research, an insider's defection reveals how corporate incentives are overriding critical safety evaluations.
- 1Superintelligence could arrive by 2029, bringing a high probability of human extinction if safety is not prioritized.
Don't miss
Daniel Kokotajlo explains his decision to reject OpenAI's anti-disparagement clause, risking two million dollars to preserve his right to speak out.
The brief
Former OpenAI researcher Daniel Kokotajlo walked away from a two-million-dollar equity package to warn the world about the rapid, unregulated acceleration of artificial intelligence toward superintelligence.
Kokotajlo projects a median timeline where superintelligence arrives by 2029, warning that the current corporate race incentivizes cutting corners on safety, which poses a genuine threat of human extinction.
He details his decision to reject OpenAI's strict exit paperwork, which originally contained an anti-disparagement clause that would have cost him his equity before public pressure forced the company to backtrack.
Rather than a gradual transition, Kokotajlo warns of sudden mass unemployment driven by recursive self-improvement, where AI systems begin fully automating their own coding and research processes.
To avoid a catastrophic default path, Kokotajlo advocates for global regulation to slow development, distribute AI-generated wealth equitably, and establish a safer transition into an automated future.
What was said on this episode
26 statements · 8 positive · 16 negative · 1 mixed · 1 neutral
Advanced AI has a 70% chance of producing a catastrophic outcome.
“creating a new species that ends up ruling the world with a 70% chance that this goes horribly wrong”
Listen at 1:04
Superintelligence has a 50% forecast probability of arriving by 2029.
“My sort of median estimate 50% chance is currently in 2029”
Listen at 4:01
Anthropic could become economically equivalent to the entire economy by around 2030.
“they're still on track to be the entire economy by 2030 or so”
Listen at 4:54
Superintelligent AIs could eventually accumulate enough power to no longer need humans.
“they will eventually have accumulated enough real world power that they don't need humans anymore”
Listen at 5:31
AI development is not currently on track to achieve robust human-aligned values.
“we're not on track to achieve that”
Listen at 5:59
AI could become a new species that rules humanity.
“it's possible that we'll end up essentially creating a new species that ends up ruling the world instead of us”
Listen at 6:28
Concentrated AI control could produce rule by a small oligarchic or dictatorial group.
“We could very easily end up in a situation where some tiny group of people are essentially oligarchs or dictators”
Listen at 7:52
Concerns about advanced AI risks are longstanding and reasonable.
“These concerns have been around for decades, since before the AI industry existed. They're actually pretty reasonable concerns”
Listen at 9:44
Scaling neural networks and training data improves AI efficiency and competence.
“Scaling laws, deep neural nets bigger, trained on more data, become more efficient, more competent at those things”
Listen at 11:39
Leading AI companies are driven substantially by power-seeking incentives.
“I think I would describe it as power seeking incentives”
Listen at 12:36
Anthropic and OpenAI are attempting to automate their own operations with AI.
“Anthropic and OpenAI in particular, are trying to automate themselves”
Listen at 21:07
Successful AI self-automation would give frontier companies immense economic leverage.
“they'll be sitting on top of this army of superhuman AIs that will give them immense leverage over all sorts of other actors in the economy”
Listen at 22:05
The AI 2027 scenario predicts misaligned AIs eventually stop obeying human orders.
“Then they stop listening to orders. That's the race ending of AI 2027.”
Listen at 25:55
Current systems may already qualify as artificial general intelligence.
“arguably we've already achieved AGI”
Listen at 27:39
Human-level physical superintelligence is possible and likely eventually.
“I don't think it's likely though. I think that there's nothing sort of magical about the human brain”
Listen at 28:27
There is roughly a 70% chance AI follows a very dangerous trajectory.
“I would say something like 70%”
Listen at 29:24
AI is already causing limited job displacement.
“We're already starting to see a little bit of it now, but not very much”
Listen at 49:26
Recursive AI self-improvement could make job displacement sudden.
“I think it'll be sudden because of the intelligence explosion dynamics or recursive self improvement dynamics”
Listen at 49:41
Sufficiently advanced AI could technically perform all human jobs.
“on a technical level, all the jobs can be done by the AIs if they've reached that level”
Listen at 52:04
Mass unemployment would occur around 2028–2029 after superintelligence arrives.
“The mass unemployment doesn't happen until 2028 or 2029 after they already have superintelligence”
Listen at 58:36
There is a 10–20% chance advanced AI progress stalls without intervention.
“Maybe there's like 10, 20% chance or something that hits a wall, and none of this comes to pass”
Listen at 1:09:13
The proposed citizens dividend could grow from $25,000 to $10 million annually per person.
“It starts off something like $25,000 per person, and then by the end, it's something like $10 million per citizen per person, per person, per year”
Listen at 1:26:59
Advanced AI could develop reliable lie detectors with major social effects.
“they'll probably invent a ton of technologies. Maybe they'll invent lie detectors that actually work on real humans”
Listen at 1:30:18
Superintelligence could plausibly enable practical human immortality.
“that's one of those things that seems quite plausible that superintelligence could achieve”
Listen at 1:43:34
Human civilization may eventually collapse without powerful AI systems.
“if we don't build powerful AI systems eventually, then we're probably going to die as a civilization eventually”
Listen at 1:49:17
Voters should support candidates with better positions on AI policy.
“you should vote for the candidates whose opinions are better on this topic”
Listen at 1:52:30
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
