Claims

Factual · Their own work or organisation

“MIRI has something called the Visible Thoughts Project, which did not get enough funding and enough personnel and was going too slowly.”

Eliezer Yudkowsky, MIRI founder, interviewee · 6 April 2023 · Eliezer Yudkowsky - Why AI Will Kill Us, Aligning LLMs, Nature of Intelligence, SciFi, & Rationality (Dwarkesh Podcast)

MIRI ran a Visible Thoughts Project that did not get enough funding or personnel and was progressing too slowly.

Mixed Partly true and partly false in roughly equal measure.

The evidence is the person's or their organisation's own records.

Said for an organisation they lead or work for; it counts on both records.

What would count as true or false

Judge as of 2023-04-06. The claim has three parts: (a) MIRI ran a Visible Thoughts Project; (b) it lacked enough funding or personnel; (c) it was progressing too slowly. True: MIRI announcements, LessWrong or MIRI blog posts, or contemporaneous reporting confirm (a), and show the project was underfunded or understaffed and slow-moving, e.g. it fell short of its dataset targets or timeline or was wound down for those reasons. Mostly true: (a) holds and the project clearly stalled or fell short, but the documented reason differs somewhat in detail, e.g. writer-recruitment difficulty rather than money. Mixed: the project existed, but only one of the problems (resources, slow progress) is supported. Mostly false: the project existed, but the record shows it was adequately resourced and on schedule, or ended for unrelated reasons. False: no such MIRI project existed.

Evidence

1. Visible Thoughts Project and Bounty Announcement - Machine Intelligence Research Institute Machine Intelligence Research Institute (MIRI) · 29 November 2021 Own record

   “In this case, we tried to execute this project in a closed way in mid-2021, but work was not proceeding fast enough.”

2. Visible Thoughts Project and Bounty Announcement - Machine Intelligence Research Institute Machine Intelligence Research Institute (MIRI) · 29 November 2021 Own record

   “We have $200,000 in prizes for building the first fragments of the dataset, plus an additional $1M prize/budget for anyone who demonstrates the ability to build a larger dataset at scale.”

3. I don't think MIRI "gave up" — LessWrong LessWrong · 3 February 2023 Own record

   “Launched the Visible Thoughts Project, which while not a stunning exemplar of efficient success *has* produced two complete 1000-step runs so far, with more in the works”

4. MIRI's May 2024 Newsletter — LessWrong LessWrong (linkpost of intelligence.org, by MIRI staff member Harlan) · 15 May 2024 Own record

   “Neither hope played out, and in the intervening time, the ML field has moved on.”

Reasoning

MIRI's November 2021 announcement confirms that it ran the Visible Thoughts Project. It also says that in an earlier closed attempt in mid-2021 "work was not proceeding fast enough" (E1). A 2023 LessWrong post calls the project "not a stunning exemplar of efficient success" and says it had produced two complete 1000-step runs so far, with more in the works (E3). That supports the claim that progress was slow. However, none of the quotes shows a shortage of funding or staff. E2 shows MIRI offering $200,000 in prizes plus a $1M prize/budget, but does not say whether the project had enough funding or staff. Only the slow-progress part is supported, which fits the "mixed" verdict.

Cite this claim

Eliezer Yudkowsky, “MIRI ran a Visible Thoughts Project that did not get enough funding or personnel and was progressing too slowly.” (said 6 April 2023). Verdict: mixed. Reliability API, claim c_dca2cbb4c1f7, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_dca2cbb4c1f7