London, we’re taking the next step! 🚙 We’re officially beginning autonomous driving with a trained specialist behind the wheel. We can’t wait to offer Londoners a quiet, convenient, and magical way to connect to the Tube, bus, or their final destination later this year.
I got a comprehensive update on 'mech interp' from @NeelNanda5 at Google DeepMind.
Neel helped make reading AI minds into a thriving field of ML.
But he has had a change of heart: it's not the silver bullet he once hoped and many others still believe it to be.
Still, they've had some big successes understanding what AIs are really thinking, and Neel thinks pairing those tools with other approaches to get 'defence in depth' remains our best and only option when deploying superhuman AI models.
Neel and I tried to cover most of what you'd want to know be up to date on this whole topic:
9:50 How Neel changed his mind on mech interp
16:00 The biggest successes so far
20:13 Probes are great
29:30 Why it won't solve all our problems
40:38 Interpretability can't reliably find deceptive AI
53:17 'Self-preservation' isn't always what it seems
1:02:25 Will AIs learn to lie in their chain of thought?
1:17:14 Models can tell when they’re being tested and act differently
1:38:24 Why everyone's excited about sparse autoencoders (SAEs)
1:47:55 Why SAEs aren't so great
2:13:11 Lessons from the mech interp hype
2:27:29 Neel’s new research philosophy
2:39:42 Who should join the mech interp field
Enjoy! Links below.
🧵SWP by 1086 Gabriel Ahlfeldt (LSE), Nikodem Szumil (UCL) + Jagdish Tripathy (BoE) quantifies the impact of house-price growth on mortgage demand in the UK via the “housing-consumption channel”… https://t.co/iqQtfrx7H0
I have the opportunity to bring some #datascience researchers into the public sector to do 1 week research visits w/ seminar. I would love you to get in touch if you or someone you know fits the bill. Helpful but not essential for work to be econ/social science related #ML#AI
Last night, BSE Research Professor Jaume Ventura received the 2022 National Research Prize of Catalonia "for his contributions to Macroeconomics and International Economics." 🏆
Join Professor of Financial Economics Tarun Ramadorai [@TRamadorai] for his inaugural lecture @imperialcollege titled 'Household decisions in financial markets: the structure beneath the surface'
Register for online or in-person tickets here ⬇️
https://t.co/KyDqINQHBk
This paper by L. Zingales on how to prevent economists' capture (https://t.co/fSuqFBfOLI) and his associated column on US economists' research on Uber (https://t.co/kkgPC3aTyP) are balanced, solution-oriented & really worth reading
Great question. I have found a few types of mentorship useful.
A lot is similar to what a grad advisor does, but just because we got a PhD doesn’t mean we’re fully-baked!
It may sound like a lot but not all of these “services” have to be by the same person…1/N
Great workshop on household finances and housing yesterday at the @bankofengland https://t.co/M4Seebin8Q
I learned a lot from outstanding presentations & discussions. Thanks for the opportunity to present joint work w @AlinaBartscher @MSchularick & Ulrike Steins
An example of the complacency that @SoumayaKeynes writes about: with a fundamental macroeconomic problem of economic growth, the @UKRI_News and @ESRC have awarded £0 in funding to research in macroeconomics in the last many years.
I was told last night that this paper with @eduardoyeyati and Nathan Converse was accepted at the @RevOfFinStudies! So happy about this outcome. Will do a 🧵 on the process and results soon, but now it’s time to celebrate!!
@arpitrage@johnjhorton If anyone on this thread is looking for resources to brush up on their Python, here's a criminally shameless plug for 'Coding for Economists':
https://t.co/2sf2oSdsEu
Please tell me what's not clear, what's missing, and what could be improved—contributions welcome!