Eric Neyman (@ericneyman) 's Twitter Profile
Eric Neyman

@ericneyman

Professional reference class tennis player. I like non-fillet frozen fish, packaged medicaments, and other oily seeds.

ID: 1520538457

linkhttp://ericneyman.wordpress.com calendar_today15-06-2013 21:54:23

2,2K Tweet

1,1K Takipçi

120 Takip Edilen

Eric Neyman (@ericneyman) 's Twitter Profile Photo

Eh, I think it's neither. Instead: - In certain cases, precision of language. Sometimes, "my prior is" is more precise than "I think that". - And then people get kinda sloppy and use it in places where "I think that" would work just as well, but without an intention to signal.

Eric Neyman (@ericneyman) 's Twitter Profile Photo

I was at the SF Exploratorium yesterday, which featured an Anthropic-sponsored AI exhibition. It had an exhibit that downplays AI x-risk by comparing x-risk fears to fears about earlier technologies. Anthropic are you guys aware that this exhibit is being run in your name??

I was at the SF Exploratorium yesterday, which featured an Anthropic-sponsored AI exhibition. It had an exhibit that downplays AI x-risk by comparing x-risk fears to fears about earlier technologies.

<a href="/AnthropicAI/">Anthropic</a> are you guys aware that this exhibit is being run in your name??
Eric Neyman (@ericneyman) 's Twitter Profile Photo

Wise words from Vitalik, as usual. Though it makes me wonder where to draw the line. I'm glad we banned slavery, for instance, rather than ending up with a "live and let live" compromise between northern and southern states. And what about factory farming?

Jesse 🔸⏹️ (@politicalkiwi) 's Twitter Profile Photo

I am an effective altruist and I eat meat although I think it's morally better to be vegan. Do any EAs out there have recommendations for good animal welfare charities to donate to, to offset my consumption?

Owain Evans (@owainevans_uk) 's Twitter Profile Photo

New paper: We trained GPT-4.1 to exploit metrics (reward hack) on harmless tasks like poetry or reviews. Surprisingly, it became misaligned, encouraging harm & resisting shutdown This is concerning as reward hacking arises in frontier models. 🧵

New paper:
We trained GPT-4.1 to exploit metrics (reward hack) on harmless tasks like poetry or reviews.
Surprisingly, it became misaligned, encouraging harm &amp; resisting shutdown
This is concerning as reward hacking arises in frontier models. 🧵
Forecasting Research Institute (@research_fri) 's Twitter Profile Photo

Respondents—especially superforecasters—underestimated AI progress. Participants predicted the state-of-the-art accuracy of ML models on the MATH, MMLU, and QuaLITY benchmarks by June 2025. Domain experts assigned probabilities of 21.4%, 25%, and 43.5% to the achieved outcomes.

Respondents—especially superforecasters—underestimated AI progress.

Participants predicted the state-of-the-art accuracy of ML models on the MATH, MMLU, and QuaLITY benchmarks by June 2025.

Domain experts assigned probabilities of 21.4%, 25%, and 43.5% to the achieved outcomes.
Eric Neyman (@ericneyman) 's Twitter Profile Photo

I remember when I was 8, my parents and I went to a buffet that was cheaper for kids 7 and under. My mom asked me if they could say that I'm 7 and I said no