måndag 22 juni 2020

My conversation with James Miller on the Swedish covid-19 strategy

There has been much talk the last few days about the interview that Swedish Minister for Foreign Affairs Ann Linde did with German broadcaster Deutsche Welle about the Swedish covid-19 strategy. The combination of arrogance and ignorance she displayed made the interview painful to watch.

Last Thursday American economist James Miller and I recorded an episode of his podcast Future Strategist on the same topic. The episode can now be heard on Soundcloud or Poddtoppen or whatever is your favorite pod platform. I tried to give a fair and level-headed account of the Swedish approach to covid-19 and how it differs from those of other countries, but I'll leave it to the listener to judge whether I've mitigated the damage to Sweden's international reputation done by Linde, or if I exasperated the situation even further. Either way, the sound quality was unfortunately not great.

lördag 20 juni 2020

Academic freedom and the case of Stephen Hsu

Since reading the excellent book The Coddling of the American Mind by Greg Lukianoff and Jonathan Haidt last year, I have become more acutely aware of how increasingly bad the situation is becoming for academic freedom in the United States. We see similar tendencies in Sweden, but the situation across the Atlantic appears worse.

Further incidents after the publication of Lukianoff and Haidt underline how bad the situation is. A few days ago I learned (via Slate Star Codex) about the case of Stephen Hsu, a physics professor at Michigan State University, who was also (until yesterday) their VP in charge of research and graduate studies. Since June 10, he has been that target of an angry mob (starting with a Twitter thread from the Michigan State Graduate Employees Union) calling for his removal as VP, based on unsubstantiated and (frankly speaking) ridiculous charges of racism.

I am now reached by the highly distressing news that the president of Michigan State University has caved in to the mob and asked Hsu to resign. Hsu complied, albeit under protest. This is terrible. Academic freedom is under assault, and (to borrow Hsu's own words) "Academics and Scientists must not submit to mob rule".

måndag 8 juni 2020

EA and me

I was recently asked to contribute a 150-word text about my relation to effective altruism (EA), for publication in a pamphlet-of-sorts along with similar texts from other supporters of EA. I submitted the text below. After a couple of rounds of email exchange with the editor about how my text could be made less controversial we came to a standstill, and the text was rejected. I nevertheless like it, so am happy to share it here, with URL links added for your convenience:
    About a decade ago I began my mid-career shift from being a typical ivory tower mathematics professor towards an increasing focus on AI safety, existential risk and related topics. This move was largely driven by the EA-like idea of wanting to address the world's most pressing issues. Yet the EA movement appeared on my radar only gradually. Initially I was skeptical, as ranking ways to do good in terms of efficiency reminded me of Bjorn Lomborg’s ploy that we shouldn’t fight climate change because fighting malaria is more cost effective. Of course there are many things we ought to do! But when in 2017 I read Will MacAskill’s Doing Good Better, I realized that current EA thinking is less naive than I thought. I have since then become increasingly impressed by all the good theoretical and practical work done in EA, and I now consider myself a warm supporter of both the movement and its core ideas.

onsdag 3 juni 2020

Is GPT-3 one more step towards artificial general intelligence?

How does the pandemic affect the long-distance race between (and now I borrow Nick Bostrom's metaphor) the galloping stallion which is humanity's technological capability, and the foal on unsteady legs which is humanity's wisdom? I don't know, but it does seem that AI development is the kind of activity that can move on pretty much unhampered by lockdowns etc. Last week, OpenAI announced their stunning natural language processor GPT-3, which is an enormously scaled-up sequel to their amazing and much-discussed GPT-2 last year. Yannic Kilcher at ETH has gone over their paper carefully and offers his insights in a very instructive video lecture:

The attentive viewer will notice that Kilcher, while accepting that GPT-3 does a bunch of impressive-looking things, what it does is not so much reasoning as pattern matching and copy-paste. This can be interpreted as Kilcher leaning in the direction of a negative answer to the question in the title of the present blog post. I am personally somewhat more inclined towards a positive answer, as is prominent tech blogger gwern:
    GPT-3 is scary because it’s a tiny model compared to what’s possible, with a simple uniform architecture trained in the dumbest way possible (prediction of next text token) on a single impoverished modality (random Internet text dumps) on tiny data (fits on a laptop), and yet, the first version already manifests crazy runtime meta-learning - and the scaling curves still are not bending! [...] In 2010, who would have predicted these enormous models would just develop all these capabilities spontaneously, aside from a few diehard connectionists written off as willfully-deluded old-school fanatics by the rest of the AI community? [...] GPT-3 is hamstrung by its training & data, but just simply training a big model on a lot of data induces meta-learning without even the slightest bit of meta-learning architecture being built in; and in general, training on more and harder tasks creates ever more human-like performance, generalization, and robustness.
The semi-fictitional dialogue between an anonymous machine learning researcher NN, and Scott Alexander, in connection with last year's release of GPT-2, is worth recalling again:
    NN: I still think GPT-2 is a brute-force statistical pattern matcher which blends up the internet and gives you back a slightly unappetizing slurry of it when asked.

    SA: Yeah, well, your mom is a brute-force statistical pattern matcher which blends up the internet and gives you back a slightly unappetizing slurry of it when asked.

tisdag 19 maj 2020

My conversation with Jaan Tallinn on AI risk and AI ethics in the age of covid-19

Today, as part of the AI ethics seminar series at Chalmers, I had a conversation with Jaan Tallinn that was broadcast live on Zoom. The topic of the day was AI risk and AI ethics in the age of covid-19, and a live recording of the discussion is available here.

One topic that was mentioned in our conversation is the (possibly crucial) role of embedded agency in AI alignment. I will have more to say about that in another live talk, on Wednesday next week, May 27, at 14:00 CEST. It's a Svenska statistikfrämjandet event, to which you can register here, and the title of my talk is AI alignment, embedded agency and decision theory.

torsdag 7 maj 2020

Fortsatt debatt i DN om coronastrategier

Som tidigare meddelats här på bloggen publicerade jag tillsammans med Olof Johansson Stenman, Joacim Rocklöv, Stefan Schubert och Markus Stoor på Valborgsmässoafton en debattartikel i DN rubricerad Alternativ coronastrategi för Sverige kan rädda liv. I måndags kom Martin Kulldorffs replik Hög dödlighet idag kan ge färre döda på sikt, och igår publicerades vår slutreplik Låt en kommission utvärdera svenska coronastrategin, där vi utöver att svara Kulldorff passade på att anknyta till ett par andra reaktioner på vår ursprungsartikel: en TT-intervju med Anders Tegnell, och en ledarkrönika i DN av Lars Calmfors. Så här inleds vår slutreplik:
    Martin Kulldorff argumenterar i sin replik för att Sverige även i fortsättningen endast bör bromsa Covid-19, och inte försöka stoppa den: vad vi i vår artikel kallade "bromsstrategin". Vi välkomnar hans replik varmt eftersom huvudbudskapet i vår debattartikel var just att det behövs mer öppen diskussion om fördelar och nackdelar med olika coronastrategier.

    Man kan diskutera hur de olika strategierna ser ut och vad de bör kallas. Kulldorff menar att den svenska strategin bör kallas "åldersstrategin" eftersom ett effektivt skydd av de äldre minskar konsekvenserna av den smittspridning som krävs för att nå flockimmunitet. Vi håller med om att det är klokt att särskilt skydda de äldre. Men valet kvarstår mellan att enbart bromsa smittan till nivåer sjukvården klarar av ("bromsstrategin") och att snabbare försöka få ner den till lägre nivåer, som till exempel Norge har gjort ("hammaren och dansen").

    Här är det viktigt att skilja mellan övergripande strategi, å ena sidan, och konkreta åtgärder med vars hjälp man implementerar den strategin, å den andra. "Hammaren och dansen" är en övergripande strategi, som syftar till att först få ner smittan till låga nivåer, och sedan se till att den hålls nere eller trycks tillbaka ytterligare.

    För att genomföra denna strategi behövs...

Läs den spännande fortsättningen här!

torsdag 30 april 2020

På DN Debatt om det fundamentala vägvalet i svensk coronastrategi

Jag har åter begått debattartikel i ämnet covid-19. Tillsammans med Olof Johansson Stenman, Joacim Rocklöv, Stefan Schubert och Markus Stoor försöker vi dra uppmärksamheten till det fundamentala vägvalet i svensk coronastrategi. Artikeln fick rubriken Alternativ coronastrategi för Sverige kan rädda liv, och inleds på följande vis:
    Den svenska corona­debatten handlar till stor del om frågor som Folkhälsomyndighetens kommunikation, huruvida intensivvården klarar trycket och skillnader mellan Sverige och andra länder i åtgärder för fysisk distansering. De är alla viktiga frågor, men vi menar att debatten nästan helt missat det allra viktigaste: ett mer fundamentalt strategival som sällan berörs.

    Hur ska man hantera farliga och smittsamma sjukdomar? Ett sätt är att utrota sjukdomen, som skedde med exempelvis smittkoppor. Efter framgångsrik bekämpning kunde den förklaras utrotad 1980. Vad gäller covid-19 hade utrotning kanske varit möjlig på ett tidigt stadium av den inhemskt kinesiska spridningen, men är i nuläget knappast realistiskt.

    De alternativ som därmed återstår är huvudsakligen två: å ena sidan vad man kan kalla "bromsstrategin", och å andra sidan det som i den internationella debatten har kommit att kallas "hammaren och dansen" (i den vetenskapliga litteraturen benämns de oftast "mitigation" respektive "suppression").

    Folkhälsomyndigheten bestämde sig tidigt för "bromsstrategin". Tanken är att epidemin bara kan hejdas genom att en tillräcklig andel av befolkningen smittas för att uppnå så kallad flock­immunitet. För covid-19 är den vanligaste bedömningen att denna andel är omkring 60 procent. Utmaningen ligger då i att se till att smittan sveper genom befolkningen tillräckligt långsamt och kontrollerat för att sjukvården inte ska överlastas. Smittan ska inte stoppas, endast bromsas.

    "Hammaren och dansen" är en mer ambitiös strategi. I inledningsfasen - Hammaren - inför man...

Med denna lilla cliffhanger uppmanar jag er att läsa fortsättningen i DN, och jag hoppas att chefredaktör Wolodarski fortsatt lever upp till sitt löfte om slopad betalvägg för coronaartiklar.