Tor Barstad

[Tor Barstad @Tor_Barstad]

@seymourtherat @McJuggerNuggets I would say that deliberately not having children with one's cousins is a more clear instance of eugenics, since it helps the gene pool long term.

Meanwhile people with Down Syndrome rarely have children.

[Tor Barstad @Tor_Barstad]

@deathtouch2k Joined September 2015?
This is your only tweet?

[Tor Barstad @Tor_Barstad]

> Hvordan netto-effektene av forskjellige former for bistand er

Et sidespor, litt sånn apropos:

Penger til bevaring av regnskog anser jeg som potensielt ganske ille fra et dyrevelferds-perspektiv.

Tror jeg heller ville bli født som et tilfeldig dyr på en gård enn et tilfeldig dyr i regnskogen.

[Tor Barstad @Tor_Barstad]

Mitt inntrykk er at dette fondet står bak relativt effektiv/målbar bistand. Skal ikke si det med 100% sikkerhet, men det er mitt inntrykk.

Hvordan netto-effektene av forskjellige former for bistand er mtp "hvor mye mer/mindre velvære/lidelse får vi alt i alt?" er jo vanskelig å tenke på uten å være ganske spekulativ og kynisk. Men førstehånds-effektene virker jo ihvertfall ganske positive.

[Lewis Bollard @Lewis_Bollard]

Denmark's new government just announced major pig welfare reforms.

They're banning the practices of restraining sows in farrowing crates, breeding them beyond their natural limits, and cutting off piglets' tails.

They're also requiring pigs get more space and rooting material. And they're stepping up farm inspections and penalties for cruelty.

All while the US Congress is trying to wipe out the few state laws that protect pigs via the Save Our Bacon Act buried in the farm bill.

Denmark just showed what's possible for pigs when voters demand it — and politicians listen.

[Tor Barstad @Tor_Barstad]

@lauriewired @pfau Then again, maybe I'm not familiar with the world @pfau is describing. I don't remember a past where "AI and fusion were laughable dead ends". And I guess I was fairly young in the 90s.

[Tor Barstad @Tor_Barstad]

> predictions that went wrong

Hmm. I also notice @pfau saying "Predictions are hard", perhaps with a connotation that people who were excited about nanotech/biotech were strongly indicated to have been wrong somehow?

I don't share this sense.

I feel like people often have mentioned "Drexler-like nanotechnology" as one reason for why superintelligent AI could be so transformative.

Often people have described scenarios where we obtain superintelligence and then advanced nanotech is developed (but while also allocating some probability-weight to things being more in the reverse order of that).

I am aware that there were some timed predictions also. And I also acknowledge that people who were excited about nanotech will lose some amount of "Bayes-points" the more time passes without hints of their vision coming to life. But still.

Anyway, I for one am still very much excited about advanced biotech and molecular manufacturing. I may very well be wrong to give so much credence/salience to "Drexler-like nanotechnology". But I don't feel as if this has been demonstrated to me by real-world outcomes. Not so far.

[Tor Barstad @Tor_Barstad]

@larry_88a @OlenaRohoza Strikes on Russia are not an attempt to "destroy Russia", but rather to reduce Russia's offensive capabilities. The Russian regime has long been offered the opportunity to negotiate earnestly, but apparently they need more convincing.

[Tor Barstad @Tor_Barstad]

@akarlin Do you have thoughts in regards to hardware stocks? (ASML, TSMC, Samsung, Micron Technology, etc)

Totally fine if this is not within your realm of care/consideration.

(Personally I'm bullish, but could easily end up burned, as these stocks are up a lot already)

[Tor Barstad @Tor_Barstad]

@akarlin Seems fair. If anything it underplays the gulf between English and everything else.

[Tor Barstad @Tor_Barstad]

In general, earlier parts of the USSR are poorer compared to what they would have been if they never lived under communism.

In Ukraine's case, they are also less populous (due to the Holodomor).

Did the USSR also do some cool stuff? Sure. The phrase "built for them" is a bit off-putting though, and I would guess that many of the people who worked on that metro were from present-day Ukraine.

[Tor Barstad @Tor_Barstad]

I guess one thing to note is that I use mostly Claude Code, even though it's more expensive. As a system it's similar to Codex, but different enough for me to currently have a significant preference for one over the other.

Another thing to note is that if demand at some point exceeds the total supply of compute (due to AI being good enough for the compute to give a high ROI), having the chips/data-centers could be a moat of sorts (not a moat that lasts forever, but still).

That said, I don't disagree when you predict that "nobody can charge monopoly prices".

If we take it as a pre-condition to be very bullish on AI (a very non-obvious assumption I don't expect you to share), investments in ASML/TSMC/etc could plausibly be safer than investments in OpenAI/Anthropic/etc (I know neither category is safe).

[Tor Barstad @Tor_Barstad]

> mens COOene febrilsk letter etter det minste tegn til kvantifiserbar positiv ROI på minst ett prosjekt

Slik kunne man vel også snakket med henhold til hvor mange ansatte man har, ikke kun AI-arbeid. Ofte ikke så rett frem å analysere ROI, og da spesielt ikke over en kort tidsperiode, og i henhold til et verktøy som er spredd ut heller enn fokusert på bestemte prosjekter.

Og ROI vil jo også forandre seg over tid. Hvor kapable disse AI-systemene er står jo ikke stille. For min del har jeg opplevd en ganske betydelig kvalitativ utvikling også det siste året.

Jeg merker jeg blir fascinert av hvor mye disse systemene programmerer riktig på første forsøk. De har definitivt sine tilkortkommenheter også (ikke uten grunn at de trenger å bli guidet), men merker at de klarer mye større chunks med arbeid på en gang nå enn før (og gjøre mindre feil, skriver mindre "rotete" kode, er mindre håpløse på jobbing med design, osv). Blir interessant å se hvordan disse trendene utvikler seg videre.

> datasentrene de har fått tilgang til av xAI de første to månedene, før de skal begynne å betale 1.5 mrd dollar per måned.

Ok, jeg noterer meg at discount fra xAI potensielt kan tegne et misvisende bilde for Q2 (kanskje vippe dem over mtp å fremstå profitable, heller enn at de vipper under). Visste ikke det.

Samtidig verdt å bemerke seg hvorfor de nå kjøper compute av xAI. De har selv investert i compute på forhånd, basert på gjetninger om hvor stor etterspørsel det ville være nå. Også viste det seg at etterspørselen ble større enn det de så for seg (turte å satse på).

Så de tok en kalkulert risiko hvor de satset stort på compute, men ikke stort nok. (Er slik leser jeg situasjonen ihvertfall. Særdeles godt mulig at du ikke deler den tolkningen.)

[Tor Barstad @Tor_Barstad]

@RuxandraTeslo @krishnanrohit What about now @RuxandraTeslo? What does your female intuition say?

Selling? Holding? Buying more?

(Most of my liquid assets are in TSMC, ASML, Micron Technology and Samsung btw. But always interested in becoming more decorrelated in ways that are non-boring 🤠)

[Tor Barstad @Tor_Barstad]

@AaireVoltaire @christianjbdev I don't think the grocery store chains have margins that are particularly large.

[Tor Barstad @Tor_Barstad]

As a Norwegian, I tend to go to order food online much more ofte than I go to the grocery store.

Main thing I miss is being able to order greens (cucumber, onions, bell peppers, etc) to my door that already have been cut into small bits, at a reasonable price.

I do think protectionist policies make some foods more expensive than they need to be (although maybe it's good to restrict meat imports in cases where animal welfare standards are significantly lower).

Maybe Scottish supermarkets are better, but I don't feel myself unfortunate to be missing out.

[Tor Barstad @Tor_Barstad]

Seems non-obvious to me that I should agree with your premise.

But of course, every AI company will want good PR, whether in the lead or not.

Anyway, I do try the different systems from time to time (Codex, Cursor with GPT 5.5, etc) and my anecdotal experiences so far makes me want to stick with Claude Code as primary.

Evidently I'm not alone, since so many prefer prefer Claude despite the higher price. But I note also that my experience seems to not be universal.

[Tor Barstad @Tor_Barstad]

@akarlin Then again, I would feel frustrated by getting too little beverage (no mugs or filling stations on the table), and by the water not having ice cubes.

[Tor Barstad @Tor_Barstad]

@kelseynicole780 @MickeyZhivago @ZelenskyyUa Certainly the competition is between those two countries at the moment. Russia was faster with fiber-optics. But it does seem to me that Ukraine is more ahead at the moment, thanks in part to Starlink. I do expect US/China to race ahead over time.

[Tor Barstad @Tor_Barstad]

@MickeyZhivago @ZelenskyyUa At the moment it seems that Ukraine has more advanced capacity for the use of drones than any other country in the world.

They became advanced with the help of support from other countries, but still.

Necessity is the mother of invention.