Google and OpenAI's AI product announcements over the past month have transformed the state of AI and show the breadth and pace of change
The last month has transformed the state of AI, with the pace picking up dramatically in just the last week. AI labs have unleashed a flood of new products …
Related coverage of scrutiny around Google’s AI Overviews shows the constraint on that race. Fast rollout can widen availability and visibility, but product quality and user trust remain central to whether new AI features stick.
First-order effects
Google and OpenAI become more directly comparable across a broader set of AI products, raising the immediate pressure to match each other’s release cadence and product scope.
Users and developers face a faster-moving choice set, increasing the value of testing products on real tasks rather than treating any one release as a stable endpoint.
Second-order effects
The compressed launch cycle pushes competing AI labs to emphasize distribution, integrations, and product packaging—not only underlying model capability—to retain attention and usage.
As AI features reach more user-facing surfaces, reliability concerns highlighted by Google’s AI Overviews coverage can become a competitive differentiator alongside speed.
Third-order effects
If repeated, this pattern shifts AI competition from episodic model launches toward continuous product iteration, where durable advantage depends on both model progress and routes to users.
The industry may become more segmented between labs that can rapidly commercialize capabilities through products and those whose advances remain less accessible to end users.
The trend: AI competition is moving toward continuous, distribution-led product releases in which model capability, usability, and trust are developed in parallel.
The thing about o3 is that my fellow academics were literally just starting to test o1 on hard problems in their field. It showed a lot or promise. Lots of interesting discussions about how to validate and incorporate it into academic work. Then o3 is announced & it is obsolete
The ARC benchmarking debate is obscuring the fact that test time compute seems to be the real deal based on experiments with o1 (I cover use cases & a Harvard/Stanford paper on medical uses in the post). o1 & Gemini 2 are precursors. o3 is likely better. https://www.oneusefulthin…
Alignment with what though? Also, this is a fancy way of saying “we filter the model output through a software layer that edits the output.” There's no “thinking”. There's no actual change to the model. This kind of framing is deliberately obfuscating and misleading. [embedd…
That being said, I still think it's worth taking a quick look at this paper, to get a sense of what OAI's alignment techniques look like. It may not be a new paradigm for the world, but it might be a new paradigm for OAI. https://assets.ctfassets.net/ ...
In the Deliberative Alignment paper, Guan et al. take a dataset of OAI policy-violating prompts, first use supervised finetuning to distill policy specifications into the generation model, then use a rating model with access to the policy specs to RLAIF the model. [image]
It's also true that Anthropic's 52b model was not reasoning-RL-trained in general, and was much, much less capable. But again, I wouldn't call it a new paradigm for alignment, when the core idea is the same as before. (Alas, the paper doesn't check if hiding the CoT helps.)
Astute observers may notice the similarities to Anthropic's Constitutional AI paper, where they ... take a dataset of harmful prompts, sft to distill the constitution into the generation model via revisions, and then RLAIF the model using a rating model w/ access to the [image]
So, what does the Deliberative Alignment paper have to say about Constitutional AI? They mention it in the related work, where Guan et al. point out that Anthropic 1) trained an intermediate reward model in their RLAIF, and 2) did not sft their models on CoTs. [image]
The CAI paper is great but this is a wrong take. There is a fundamental difference between CAI/RLAIF and deliberative alignment, which is the difference between “system 1” and “system 2” thinking. Ultimately in CAI the generative model answering the questions does not know the
OpenAI presents Deliberative Alignment: Reasoning Enables Safer Language Models Saturates many of their hardest safety evaluations and achieves a Pareto improvement on both under- and overrefusals https://openai.com/... [image]
Besides o3, today OpenAI also published a “new paradigm” for alignment - “Deliberative Alignment” - which, if I'm reading the paper correctly, is Anthropic's Constitutional AI approach straightforwardly applied to o1. [image]
Quite literally. Deliberative alignment: reasoning enables safer language models Introducing our new alignment strategy for o-series models, which are directly taught safety specifications and how to reason over them. https://openai.com/...
1/5 Excited that our paper on “deliberative alignment” came out as part of 12 days of @openai! By teaching reasoning models the text of our specifications, and how to reason about them in context, we obtain significantly better robustness while also reducing over refusals. 🧵 [ima…