one novel human idea can be worth more than enormous amounts of LLM inference-time scaling.
this is where human cognition retains a real economic comparative advantage.
Only 1 out of 6 reviewers of my two papers responded to my rebuttals at #ICLR2027@iclr_conf despite having a week to do so.
The purpose of a system is what it does.
@Fred834567 I wouldn't be more impressed by someone achieving what they did through a manual they were handed, over someone achieving the same without needing one.
>Guy convinces ER doc to put him on GI endoscopy list for cancer-risk symptoms
>Waitlist is 7-8 months
>Takes 2+ years
>Pulmonary specialist cancels request without telling him
>Drops him as a patient after guy confronts her
>Guy requests it 4 MORE TIMES through family doctor
>Denied 3 times until he starts yelling
>4th request approved
>Endoscopy finds esophageal cancer
>It’s already metastasized
>The denier lies to guy’s family doctor and claims he had been refusing the test all along
>Guy is now in palliative care
>Same pattern: PE untreated for months because no chest CT despite requests
>Also ignored 40 concussions
>Missed PE made him ineligible for cancer surgery
You can’t possibly hate healthcare workers enough.
Claude Fable 5.2 on the Bach Benchmark. One of the best chorales we’ve seen yet. This is the first time I remember seeing a third-inversion chord (m. 2) on this benchmark. There’s still a little awkwardness, such as the tenor doubling the third of the tonic in m. 9, but otherwise this is good solid work.
ICLR reviews
I feel terrible for my PhD students. They spent months developing deep expertise, only to receive reviews from people who don't understand the topic.
Top AI conferences need a qualification stage. We can no longer afford to let anyone submit w/o initial screening.
feels pretty good that a ~1y research effort is finally coming to fruition and a frontier model rates the first version of the paper as a "landmark study" in terms of contribution. sucks it didn't work out before the ICLR deadline though.
someone believing in a 2000 year old myth based on authority and tradition will also rather believe a contemporary myth based on authority and tradition
the relationship between totalitarianism/authoritarianism and religion is instrumental
@alexolegimas@GoogleDeepMind@andrewjkoh@ShaneLegg Isn’t it a bit naive to assume (in the paper) that:
1. Demand is fixed
2. There is no government/taxes
3. Lost wages aren’t counted as lost
4. Displaced workers find alternative employment
5. Compute profits are external and fixed
6. There are no regions and trade
@alexolegimas@GoogleDeepMind@andrewjkoh@ShaneLegg What about regional economies that depend on the wages (offshore) AI and agents substitute? What’s your take on how they cope with the loss of tax base and consumption? Why is the paper silent on this?
@a_sator Wie kann es legal sein Autos zu fahren die Terroristen für Attentate auf Weihnachtsmärkte verwenden können?
Wie kann es legal sein Messer zu verkaufen (..) ?
Dario Amodei was so worried about Y2K that his high school teacher still remembers it decades later.
He’s been a doomer for much longer than I thought!