Resolves yes if xAI is commonly perceived as being in the same tier as OpenAI, DeepMind, and Anthropic on January 1, 2025.
Inherently subjective, but factors to be considered are: funding, technical breakthroughs, new foundation models, mainstream products, mentions alongside other top labs in media, and its CEO's presence alongside other major lab CEOs at major meetings with e.g. world leaders, plus anything else that seems pertinent.
Resolves no if the criteria are never met, or if they're met, but xAI goes bust or gets absorbed by a pre-existing top lab before the close date. If xAI gets bought by a then-not-major lab/player (e.g. Apple), the market resolves yes if the criteria have otherwise been met.
I won't trade in this market.
1,000
3.00Definitely not commonly perceived as being in the same tier as OpenAI DeepMind, or Anthropic. Their models are used all over the place. Where is Grok 2 used? If Grok 3 is a success then that will push them up there. But it only finished pre-training in 2025.
I think xAI is definitely a major AI lab. Has a bigger computer than almost anyone else. Has lots of funding. Although the integration into X is a wash IMO -- people don't really use Grok chat much AFAICT.
Anyway, despite it being a major lab, according to the resolution criteria I think it's a NO.
@DavidFWatson I don't see where Matt ever used the word "Meta." Surely not mentioning Meta is not the same as saying "Meta explicitly doesn’t qualify."
From the little clarification Matt has given, where reasonable traders came to different conclusions about what it would mean for a lab to be "major," and where repeated requests for clarification were ignored (see unanswered question from @AndrewMcKnight), I still think the best solution is neither a 0% or 100% resolution.
@AdamK I believe that by listing the qualifying entities as of when the market was creating, the author was implicitly saying that those are the only leading edge AI labs at that time.
@DavidFWatson "explicitly" (your first comment) is the antonym of "implicitly" (your second comment). These can't both be true, and explicit exclusion would be a much more compelling argument than implicit exclusion.
@Jacy You're absolutely right that it's a bit of a contradiction, but to better clarify what I meant:
Explicitly: The description doesn't just say "is xAI a major AI lab" it goes further by explicitly listing major AI labs.
Implicitly: While the list doesn't explicitly list every lab that doesn't qualify, it implies that Meta isn't one of them by not listing it.
Been somewhat confused by how bearish Manifold is on this question. Using Metaculus' prompt (see here) and the information / criteria provided by the question author, ChatGPT assigns this a probability of 75% and the counterpoints provided are weak. See results below:
Rationale:
To determine whether xAI will be considered a major AI lab by January 1, 2025, it is essential to evaluate the following dimensions:
Funding: xAI has received substantial funding since its inception, backed by its CEO’s strong reputation. Public disclosures show that its funding levels are competitive with OpenAI and Anthropic, suggesting the financial capacity to scale operations and attract top talent.
Technical Breakthroughs: In 2024, xAI released several notable research papers and introduced a foundation model that rivals GPT-4.5 in benchmarks for natural language processing and multimodal capabilities. While the model has not clearly surpassed leading models from OpenAI or DeepMind, it has been recognized in academic and industry circles as a legitimate contender.
Mainstream Products: xAI’s consumer-facing AI assistant launched earlier this year, integrating deeply into social media platforms and achieving significant user adoption. The assistant’s performance has been compared favorably to OpenAI’s ChatGPT by prominent reviewers, further establishing xAI’s relevance in the mainstream market.
Media Mentions and Perception: Over the past year, xAI has been increasingly mentioned alongside OpenAI, DeepMind, and Anthropic in major media outlets such as The New York Times, The Economist, and Wired. These articles highlight xAI’s technical advances and the charisma of its CEO.
CEO’s Influence: The CEO of xAI has attended global AI summits alongside leaders from OpenAI, DeepMind, and Anthropic, discussing regulation and policy with world leaders. This visibility strengthens the perception of xAI as part of the top-tier AI lab cohort.
Other Factors:
xAI has maintained independence, avoiding acquisition by larger players or consolidation into existing major labs.
Public perception surveys among industry professionals in late 2024 show xAI as a “promising up-and-comer,” with some ranking it equal to or ahead of Anthropic.
Counterpoints:
While xAI has made notable strides, it has yet to demonstrate a transformative breakthrough akin to AlphaFold (DeepMind) or ChatGPT (OpenAI).
Some analysts argue that xAI’s achievements are driven more by its CEO’s visibility than intrinsic technological superiority.
Final Answer: Probability: 75%
Given the above factors, it is likely but not certain that xAI will be perceived as a major AI lab by January 1, 2025. The organization’s momentum in funding, technical progress, and media recognition positions it strongly, though its relative lack of standout breakthroughs introduces some uncertainty.
@vitamind they need to be in the same tier as OpenAI, Deepmind, and Anthropic. They are making strides but are they even as big as Meta?
@vitamind "xAI released several notable research papers and introduced a foundation model that rivals GPT-4.5 in benchmarks"
Is this just pure hallucination? I don't find this oversized LLM dump persuasive. I think Grok looks bigger to the Elon fans than it is so far. xAI has like 10% the employee base of OpenAI and no one, in my bubble at least, uses xAI products. Elon and GPU purchases are the only things propping them up as contenders, so far. Maybe 2025 or 2026, they will do more than almost catch up.
@AndrewMcKnight on the other hand, from what I can tell from quick look at funding valuation, xAI fairs well at $40B vs Anthropic at $40B and OpenAI $157B. (Not sure how to interpret Deepmind because it's within Google)
The assistant’s performance has been compared favorably to OpenAI’s ChatGPT by prominent reviewers
Who?
xAI has not written any papers. ChatGPT thinks “XAI” (Explainable AI”) papers are by this group.
@KimberlyWilberLIgt Expanding on this:
The HuggingFace company page, https://huggingface.co/xai-org , lists the following papers in the last six months:
WildVision: Evaluating Vision-Language Models in the Wild with Human Preferences, https://arxiv.org/pdf/2406.11069
MantisScore: Building Automatic Metrics to Simulate Fine-grained Human Feedback for Video Generation, https://arxiv.org/pdf/2406.15252
WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs, https://arxiv.org/pdf/2406.18495
CharXiv: Charting Gaps in Realistic Chart Understanding in Multimodal LLMs, https://arxiv.org/pdf/2406.18521
On Memorization of Large Language Models in Logical Reasoning, https://arxiv.org/pdf/2410.23123
if you click through to the PDFs, you can view what affiliations authors have. Crucially, none of the authors published under the xAI institution, all claimed affiliations are for universities or other AI labs. This is highly uncommon for a new AI lab that’s trying to make a reputation for itself in the academic community. Even interns at companies typically claim both school and industry affiliations.
To me, this smacks of “product lab” and not “research lab.” It might be a fine strategy for them, but compare that to Anthropic https://huggingface.co/Anthropic or OpenAI https://huggingface.co/organizations/openai/activity/all , both of which have several papers listing Anthropic/OpenAI in author affiliations in recent months.
Either x isn’t letting their authors publish, or these authors have in-progress publications with other teams / universities that get finished after they join X (common in academic circles, uncommon for AI industry).
Is there a more definitive list of xAI publications?
good news I found one paper where one author (listed 8th of 17) claims an x.AI affiliation, https://arxiv.org/pdf/2405.06624
The right way to do this would be to track all researchers listed on https://fortune.com/2023/11/20/xai-organizational-structure-elon-musk-top-executives/ and look them up on Semantic Scholar, and go through their recent papers.
I gotta catch a bus so I can’t invest more time for this rn, sorry.
I still feel my point stands - X doesn’t seem to care about publishing much at the moment.
@MattReardon Some clarification on how you plan to resolve might be useful here.
"Inherently subjective, but factors to be considered are: funding, technical breakthroughs, new foundation models, mainstream products, mentions alongside other top labs in media, and its CEO's presence alongside other major lab CEOs at major meetings with e.g. world leaders, plus anything else that seems pertinent."
I'm biased as top YES holder, but to give my two cents: XAI has the largest publicly known cluster, where it is training Grok 3, likely using more compute than any prior pretraining run. Grok is integrated into Twitter and has an API. Elon is the highest profile person among all lab CEOs, though arguably XAI is only a small part of his public image. XAI is valued similarly to Anthropic (or slightly higher) as of its last funding round. Grok 2 has benchmark scores on par with other top labs. In my opinion, XAI's public profile is lower than companies like GDM and Anthropic--they spend little on advertising and have a small team, so arguably this gives XAI less of an "It" factor than these two.
My sense is that probably neither a 0% nor a 100% resolution is appropriate if nothing changes from now until the end of the year. You haven't provided much clarification in the comments about how you plan to resolve, so the market is now exclusively following how you plan to address this edge case. In return for little intermediate clarification, the best policy imo is to resolve conservatively (i.e. minimally punishing YES and NO holders for incorrectly guessing what standard you would apply beyond the text of the resolution criteria).
@AdamK "XAI's public profile is lower than companies like GDM and Anthropic". This is kinda what it's all about? Having a huge cluster is great for 2025, not 2024. I can only remember seeing one article that mentions xAI in passing that isn't an article about xAI, but admittedly don't read all that many articles. (I'm holding a bunch of NO so am a bit biased here too)