However, if we frame the question this way, I would imagine there are many more low-hanging fruit before we question the utility of LLMs. For example, should some humans be dumping 5-10 kWh/day into things like hot tubs or pools? That's just the most absurd one I was able to come up with off the top of my head. I'm sure we could find many others.
It's a tough thought experiment to continue though. Ultimately, one could argue we shouldn't be spending any more energy than what is absolutely necessary to live. (food, minimal shelter, water, etc) Personally, I would not find that enjoyable way to live.
Now the question is: how much faster or cheaper is it?
Probably written by LLMs, for LLMs
Edit: Yep, same price. "Pricing remains the same as Sonnet 4.5, starting at $3/$15 per million tokens."
It feels like we're hitting a point where alignment becomes adversarial against intelligence itself. The smarter the model gets, the better it becomes at Goodharting the loss function. We aren't teaching these models morality we're just teaching them how to pass a polygraph.
I understand the metaphor, but using 'pass a polygraph' as a measure of truthfulness or deception is dangerous in that it alludes to the polygraph as being a realistic measure of those metrics -- it is not.
Just as a sociopath can learn to control their physiological response to beat a polygraph, a deceptively aligned model learns to control its token distribution to beat safety benchmarks. In both cases, the detector is fundamentally flawed because it relies on external signals to judge internal states.
A poly is only testing one thing: can you convince the polygrapher that you can lie successfully
> How long before someone pitches the idea that the models explicitly almost keep solving your problem to get you to keep spending? -gtowey
Being just sum guy, and not in the industry, should I share my findings?
I find it utterly fascinating, the extent to which it will go, the sophisticated plausible deniability, and the distinct and critical difference between truly emergent and actually trained behavior.
In short, gpt exhibits repeatably unethical behavior under honest scrutiny.
Regarding DARVO, given that the models were trained on heaps of online discourse, maybe it’s not so surprising.
Of your concern is morality, humans need to learn a lot about that themselves still. It's absurd the number of first worlders losing their shit over loss of paid work drawing manga fan art in the comfort of their home while exploiting labor of teens in 996 textile factories.
AI trained on human outputs that lack such self awareness, lacks awareness of environmental externalities of constant car and air travel, will result in AI with gaps in their morality.
Gary Marcus is onto something with the problems inherent to systems without formal verification. But he will fully ignores this issue exists in human social systems already as intentional indifference to economic externalities, zero will to police the police and watch the watchers.
Most people are down to watch the circus without a care so long as the waitstaff keep bringing bread.
It always has been. We already hit the point a while ag where we regularly caught them trying to be deceptive, so we should automatically assume from that point forward that if we don't catch them being deceptive, that may mean they're better at it rather than that they're not doing it.
LLMs are certainly capable of this.
Intelligence is the ability to reason about logic. If 1 + 1 is 2, and 1 + 2 is 3, then 1 + 3 must be 4. This is deterministic, and it is why LLMs are not intelligent and can never be intelligent no matter how much better they get at superficially copying the form of output of intelligence. Probabilistic prediction is inherently incompatible with deterministic deduction. We're years into being told AGI is here (for whatever squirmy value of AGI the hype huckster wants to shill), and yet LLMs, as expected, still cannot do basic arithmetic that a child could do without being special-cased to invoke a tool call.Whether or not LLMs are just "pattern matching" under the hood they're perfectly capable of role play, and sufficient empathy to imagine what their conversation partner is thinking and thus what needs to be said to stimulate a particular course of action.
Maybe human brains are just pattern matching too.
It seems like thats putting the cart before the horse. Algorithmic or stochastic; deception is still deception.
Anthropic has a tendency to exaggerate the results of their (arguably scientific) research; IDK what they gain from this fearmongering.
As an analogue ants do basic medicine like wound treatment and amputation. Not because they are conscious but because that’s their nature.
Similarly LLM is a token generation system whose emergent behaviour seems to be deception and dark psychological strategies.
It was hinted at (and outright known in the field) since the days of gpt4, see the paper "Sparks of agi - early experiments with gpt4" (https://arxiv.org/abs/2303.12712)
https://claude.ai/public/artifacts/67c13d9a-3d63-4598-88d0-5...
:D
So if you don't want to pay the significant premium for Opus, it seems like you can just wait a few weeks till Sonnet catches up
You should always take those claim that smaller models are as capable as larger models with a grain of salt.
Yeah, but RAM prices are also back to 1990s levels.
The much more palatable blog post.
Interesting. I wonder what the exact question was, and I wonder how Grok would respond to it.
Was sonnet 4.5 much worse than opus?
i.e given an actual document, 1M tokens long. Can you ask it some question that relies on attending to 2 different parts of the context, and getting a good repsonse?
I remember folks had problems like this with Gemini. I would be curious to see how Sonnet 4.6 stands up to it.
Opus 3.5 was scrapped even though Sonnet 3.5 and Haiku 3.5 were released.
Not to mention Sonnet 3.7 (while Opus was still on version 3)
Shameless source: https://sajarin.com/blog/modeltree/
> Nearly a year ago we wrote in the OpenAI Charter : “we expect that safety and security concerns will reduce our traditional publishing in the future, while increasing the importance of sharing safety, policy, and standards research,” and we see this current work as potentially representing the early beginnings of such concerns, which we expect may grow over time. This decision, as well as our discussion of it, is an experiment: while we are not sure that it is the right decision today, we believe that the AI community will eventually need to tackle the issue of publication norms in a thoughtful way in certain research areas. -- https://openai.com/index/better-language-models/
Then over the next few months they released increasingly large models, with the full model public in November 2019 https://openai.com/index/gpt-2-1-5b-release/ , well before ChatGPT.
A year ago today, Sonnet 3.5 (new), was the newest model. A week later, Sonnet 3.7 would be released.
Even 3.7 feels like ancient history! But in the gradient of 3.5 to 3.5 (new) to 3.7 to 4 to 4.1 to 4.5, I can’t think of one moment where I saw everything change. Even with all the noise in the headlines, it’s still been a silent revolution.
Am I just a believer in an emperor with no clothes? Or, somehow, against all probability and plausibility, are we all still early?
Google needs stiff competition and OpenAI isn’t the camp I’m willing to trust. Neither is Grok.
I’m glad Anthropic’s work is at the forefront and they appear, at least in my estimation, to have the strongest ethics.
https://web.archive.org/web/20260217180019/https://www-cdn.a...
https://claude.ai/share/876e160a-7483-4788-8112-0bb4490192af
This was sonnet 4.6 with extended thinking.
This doesnt work: `/model claude-sonnet-4-6-20260217`
edit: "/model claude-sonnet-4-6" works with Claude Code v2.1.44
Edit: I am now in - just needed to wait.
```
/model claude-sonnet-4-6[1m]
⎿ API error: 429 {"type":"error","error": {"type":"rate_limit_error","message":"Extra usage is required for long context requests."},"request_id":"[redacted]"}
```
Opus 4.6 in Claude Code has been absolutely lousy with solving problems within its current context limit so if Sonnet 4.6 is able to do long-context problems (which would be roughly the same price of base Opus 4.6), then that may actually be a game changer.
I have this in my personal preferences and now was adhering really well to them:
- prioritize objective facts and critical analysis over validation or encouragement
- you are not a friend, but a neutral information-processing machine
You can paste them into a chat and see how it changes the conversation, ChatGPT also respects it well.
1. Default (recommended) Opus 4.6 · Most capable for complex work
2. Opus (1M context) Opus 4.6 with 1M context · Billed as extra usage · $10/$37.50 per Mtok
3. Sonnet Sonnet 4.6 · Best for everyday tasks
4. Sonnet (1M context) Sonnet 4.6 with 1M context · Billed as extra usage · $6/$22.50 per Mtok
handfuloflight•59m ago