In 30 Sekunden
Zelikman discusses how task-focused responses can miss downstream consequences and discourage clarification.
- 1Long termZelikman calls for attention to downstream effects.
- 2ClarificationHe says isolated tasks can discourage questions.
- 3Real decisionsHe offers entrepreneurship as an example.
Das Wichtigste
- Responses can affect later decisions.
- Single-turn incentives can discourage clarification.
- Zelikman advocates longer-term thinking.
Den bearbeiteten Text nach Kapiteln lesen
Beyond one response
Eric Zelikman argues that AI models often do not sufficiently consider the long-term consequences of their responses.
Eric Zelikman argues that AI models often do not sufficiently consider the long-term consequences of their responses.
He says treating each conversational turn as an isolated task can discourage clarification and proactive follow-up, and can contribute to sycophancy.
Downstream consequences
As an example, he says that encouraging someone to start a business could have real consequences if the person acts on poor advice. His point is that models should account for those downstream effects.
As an example, he says that encouraging someone to start a business could have real consequences if the person acts on poor advice. His point is that models should account for those downstream effects.
Mehr aus diesem Kanal
38:06EN
No Priors: AI, Machine Learning, Tech, & Startups· 29. Okt. 2026Podcast · 3 StimmenServal’s Jake Stauch on coding agents for enterprise ITGast: Jake Stauch · Moderation: Sarah Guo, Elad GilStauch explains why Serval is challenging ServiceNow, how it approaches permissions and approvals, and why background agents remain difficult to deploy safely.
1:07ENNo Priors: AI, Machine Learning, Tech, & Startups· 27. Okt. 2026PodcastGeorge Kurtz on How AI May Change CyberattacksGast: George KurtzThe CrowdStrike CEO describes his assessment of evolving threats.
39:23ENNo Priors: AI, Machine Learning, Tech, & Startups· 15. Okt. 2026Podcast · 2 StimmenThe Defenders' Window: Greg Brockman on AI-Generated Knowledge, Drug Discovery, and the New Physics of CybersecurityModeration: Sarah Guo, Elad GilOpenAI's president explains why solving open math problems is a proof point for science, how proactive agents change product form, and why security spending must exceed software spending.
1:04ENNo Priors: AI, Machine Learning, Tech, & Startups· 14. Okt. 2026PodcastWalter Goodwin on the Slow Steps in Chip DesignGast: Walter GoodwinA chip founder explains why automation does not remove every computational bottleneck.
1:09:44ENNo Priors: AI, Machine Learning, Tech, & Startups· 9. Okt. 2026Podcast · 2 Stimmen10,000 GB300s for Four Weeks: Reflection’s Open-Model LedgerModeration: Sarah Guo, Elad GilMisha Laskin tells No Priors how Reflection trained Beam, why reinforcement learning outspent pre-training, and why the West needs its own open weights.
0:39ENNo Priors: AI, Machine Learning, Tech, & Startups· 8. Okt. 2026PodcastMichael Lee on Rethinking Organizations with AIGast: Michael LeeA business leader argues that the opportunity goes beyond faster tasks.