#78 – Danny Hernandez on forecasting and the drivers of AI progress

#78 – Danny Hernandez on forecasting and the drivers of AI progress

Companies use about 300,000 times more computation training the best AI systems today than they did in 2012 and algorithmic innovations have also made them 25 times more efficient at the same tasks.

These are the headline results of two recent papers — AI and Compute and AI and Efficiency — from the Foresight Team at OpenAI. In today's episode I spoke with one of the authors, Danny Hernandez, who joined OpenAI after helping develop better forecasting methods at Twitch and Open Philanthropy.

Danny and I talk about how to understand his team's results and what they mean (and don't mean) for how we should think about progress in AI going forward.

Links to learn more, summary and full transcript.

Debates around the future of AI can sometimes be pretty abstract and theoretical. Danny hopes that providing rigorous measurements of some of the inputs to AI progress so far can help us better understand what causes that progress, as well as ground debates about the future of AI in a better shared understanding of the field.

If this research sounds appealing, you might be interested in applying to join OpenAI's Foresight team — they're currently hiring research engineers.

In the interview, Danny and I (Arden Koehler) also discuss a range of other topics, including:
• The question of which experts to believe
• Danny's journey to working at OpenAI
• The usefulness of "decision boundaries"
• The importance of Moore's law for people who care about the long-term future
• What OpenAI's Foresight Team's findings might imply for policy
• The question whether progress in the performance of AI systems is linear
• The safety teams at OpenAI and who they're looking to hire
• One idea for finding someone to guide your learning
• The importance of hardware expertise for making a positive impact

Chapters:

  • Rob’s intro (00:00:00)
  • The interview begins (00:01:29)
  • Forecasting (00:07:11)
  • Improving the public conversation around AI (00:14:41)
  • Danny’s path to OpenAI (00:24:08)
  • Calibration training (00:27:18)
  • AI and Compute (00:45:22)
  • AI and Efficiency (01:09:22)
  • Safety teams at OpenAI (01:39:03)
  • Careers (01:49:46)
  • AI hardware as a possible path to impact (01:55:57)
  • Triggers for people’s major decisions (02:08:44)

Producer: Keiran Harris
Audio mastering: Ben Cordell
Transcriptions: Zakee Ulhaq

Episoder(325)

#85 - Mark Lynas on climate change, societal collapse & nuclear energy

#85 - Mark Lynas on climate change, societal collapse & nuclear energy

A golf-ball sized lump of uranium can deliver more than enough power to cover all of your lifetime energy use. To get the same energy from coal, you’d need 3,200 tonnes of black rock — a mass equivale...

20 Aug 20202h 8min

#84 – Shruti Rajagopalan on what India did to stop COVID-19 and how well it worked

#84 – Shruti Rajagopalan on what India did to stop COVID-19 and how well it worked

When COVID-19 struck the US, everyone was told that hand sanitizer needed to be saved for healthcare professionals, so they should just wash their hands instead. But in India, many homes lack reliable...

13 Aug 20202h 58min

#83 - Jennifer Doleac on preventing crime without police and prisons

#83 - Jennifer Doleac on preventing crime without police and prisons

The killing of George Floyd has prompted a great deal of debate over whether the US should reduce the size of its police departments. The research literature suggests that the presence of police offic...

31 Jul 20202h 23min

#82 – James Forman Jr on reducing the cruelty of the US criminal legal system

#82 – James Forman Jr on reducing the cruelty of the US criminal legal system

No democracy has ever incarcerated as many people as the United States. To get its incarceration rate down to the global average, the US would have to release 3 in 4 people in its prisons today.  The ...

27 Jul 20201h 28min

#81 - Ben Garfinkel on scrutinising classic AI risk arguments

#81 - Ben Garfinkel on scrutinising classic AI risk arguments

80,000 Hours, along with many other members of the effective altruism movement, has argued that helping to positively shape the development of artificial intelligence may be one of the best ways to ha...

9 Jul 20202h 38min

Advice on how to read our advice (Article)

Advice on how to read our advice (Article)

This is the fourth release in our new series of audio articles. If you want to read the original article or check out the links within it, you can find them here. "We’ve found that readers sometimes...

29 Jun 202015min

#80 – Stuart Russell on why our approach to AI is broken and how to fix it

#80 – Stuart Russell on why our approach to AI is broken and how to fix it

Stuart Russell, Professor at UC Berkeley and co-author of the most popular AI textbook, thinks the way we approach machine learning today is fundamentally flawed. In his new book, Human Compatible, he...

22 Jun 20202h 13min

What anonymous contributors think about important life and career questions (Article)

What anonymous contributors think about important life and career questions (Article)

Today we’re launching the final entry of our ‘anonymous answers' series on the website. It features answers to 23 different questions including “How have you seen talented people fail in their work?...

5 Jun 202037min

Populært innen Fakta

fastlegen
dine-penger-pengeradet
relasjonspodden-med-dora-thorhallsdottir-kjersti-idem
treningspodden
foreldreradet
rss-strid-de-norske-borgerkrigene
jakt-og-fiskepodden
rss-sunn-okonomi
takk-og-lov-med-anine-kierulf
merry-quizmas
sinnsyn
rss-kunsten-a-leve
lederskap-nhhs-podkast-om-ledelse
gravid-uke-for-uke
smart-forklart
hverdagspsyken
level-up-med-anniken-binz
hagespiren-podcast
fryktlos
rss-kull