Grok3 Launch [video]

There's also some strange wordings like "back-pocket tests."

It's 100% LLM generated.

What is much scarier is that those "quick reply" blurbs on Android/Gmail (and iOS?) will be able to be trained on your entire e-mail and WhatsApp history. That model will have your writing mannerisms and even be a stochastic mimic of your reasoning. So, you won't be able to even realize a model answered you, not a real person. And the initial message the model is responding to might be written by the other person's personal model.

The future of digital interactions might have some sort of cryptographic signing guaranteeing you're talking to a human being, perhaps even with blocked copy-pasting (or well, that part of the text shows up as unverified) and cheat detection.

Going even a layer deeper / more meta: what does it ultimately matter? We humans yearn for connection, but for some reason that connection only feels genuine with another human. Whereas, what is the difference between a human typing a message to you, a human inhabiting a robot body, a model typing a message to you, and a model inhabiting a robot body, if they can all give you unique interactions?

There is no being "on par" in this space. Model providers are still mostly optimising for a handful of benchmarks / goals, like we can already see that Grok 3 is doing incredibly well on human preference (LM Arena) however with Style Control, it's suddenly behind ChatGPT-4o-latest and Gemini 2.0 is out the picture. So even within a single domain, goal, benchmark—it's not as straightforward as to say that one model is "on par" with another.

> shouldn't we expect that anybody with the computer power is capable to compete with o1-pro?

Not necessarily. I know it may be tempting to think that Grok 3 is entirely a result of xAI having lots of "computer power", but you have to recognise that this mindset is coming from a place of ignorance, not wisdom. Moreover, it doesn't even pass off as "cynical" view, because it's common knowledge that model training is really, really complicated. DeepSeek results are note-worthy, and really influential in some respects, but it hasn't magically "solved" training, or made training necessarily easier / less expensive for the interested parties. They never shared the low-level performance improvements, just model weights and lots of insight. For talented researchers, this is valuable, of course, but it's not like "anybody" could easily benefit from it in their training regimes.

Update: RFT (contra SFT) is becoming really popular with service providers, and it's not been "standardised" beyond whatever reproductions to have emerged in the weeks prior, moreover R1 cost is still pretty high[1] at something like $7/Mtok, & bandwidth is really not great. Consider something like Google Vertex AI's batch pricing for Gemini 1.5 Pro and Gemini 2.0 Flash models, which is at 50% discount, and their prompt caching which is at 75% discount. R1 is still got a way to go.

[1]: https://openrouter.ai/deepseek/deepseek-r1/providers?sort=th...