Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

openai.com

Post reply on HN
Print view View original

221–230 of 281 posts

Re: Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

#221

Post by virgilp »

quantumspandex wrote: post #202
PunchTornado wrote: post #200

Earlier quoted context omitted.

I don't understand. do you think that people can't figure it out if they speak with llms or not from a few turns?

We don't train current LLMs to mimic the average human's writing style. We train it to be smart, helpful, and knowledgable, too knowledgeable for a human. We can easily train a LLM to pass the Turing test if we wanted to, but then it would just sound dumb or biased.

> We can easily train a LLM to pass the Turing test if we wanted to, but then it would just sound dumb or biased.

Interesting idea. "This is not dumb and biased enough, probably not a human".

Re: Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

#222

Post by pixl97 »

khalic wrote: post #191
tristanj wrote: post #185

Earlier quoted context omitted.

> Now there is no sign of actual AGI happening any time soon Do you genuinely hold this position, or do you not realize how far the goalposts have shifted? In 2022, prominent AI critic Gary Marcus offered to bet $100,000 that we wouldn't have AGI by 2029. https://garymarcus.substack.com/p/dear-elon-musk-here-are-fi... Because the definition of AGI is unclear, he defined that AGI would be achieved if an AI model could…

That’s his definition and in no way a universal one. In the meantime, it’s still very easy to differentiate between an AI and a human in a chat. You just need to know the quirks of these systems. Like counting letters, hitting the safeguards, etc. So call me when one of them can pass the Turing test against me and then we can talk about AGI

There is no accepted definition of intelligence that is usable for classification of intelligence across the sciences.

On top of that for your strong feelings, you don't have the conviction to write down a strong definition of intelligence yourself, which allows you to accelerate the goal posts up to light speed. The fun thing about writing out a formal definition is suddenly almost everything or almost nothing, including a lot of humans, has intelligence.

Not basing intelligence on your feelings of the moment makes it a hard thing to define across everything intelligence applies to.

Re: Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

#223

Post by guelo »

quantumspandex wrote: post #202
PunchTornado wrote: post #200

Earlier quoted context omitted.

I don't understand. do you think that people can't figure it out if they speak with llms or not from a few turns?

We don't train current LLMs to mimic the average human's writing style. We train it to be smart, helpful, and knowledgable, too knowledgeable for a human. We can easily train a LLM to pass the Turing test if we wanted to, but then it would just sound dumb or biased.

There is tons of money trying to get customer support bots to sound human but they fool nobody.

Re: Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

#224

Post by aleph_minus_one »

theptip wrote: post #143
stymaar wrote: post #71

Earlier quoted context omitted.

Ads work when you can server $.01 worth of ads to a user for $.0001 of server cost. I fail to see how you can make it work when doing LLM inference which is significantly costlier than web search.

OpenAI can very plausibly target ads far better than Google can. Email is a good window but lots of people talk in more depth with a chat bot. Search is a good “purchase intent moment” but people are telling a chat bot exactly what they care about in a purchase, rather than making indirect search terms and reading review sites.

> Email is a good window but lots of people talk in more depth with a chat bot.

> Search is a good “purchase intent moment” but people are telling a chat bot exactly what they care about in a purchase, rather than making indirect search terms and reading review sites.

When I look at my history with AI chatbots, 90 % of the conversations revolve around research-level problems in math and computer science. :-D

Re: Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

#225

Post by BedVibe_Studios »

ilaksh wrote: post #23

> Our mission is to ensure that artificial general intelligence benefits all of humanity. We’re introducing updates to ChatGPT that improve everyday conversations while expanding access for Free users. This clearly implies that they believe ChatGPT models are AGI and are now willing to say it out loud. Which I think is a fair interpretation of the term. They are general purpose intelligence in that you can get help f…

I found answers from free tier models to be harmful, so helping humans in general isn't what I actually noticed here.

Re: Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

#226

Post by aleph_minus_one »

kristofferR wrote: post #85
stymaar wrote: post #71

Earlier quoted context omitted.

Ads work when you can server $.01 worth of ads to a user for $.0001 of server cost. I fail to see how you can make it work when doing LLM inference which is significantly costlier than web search.

90+% of queries are probably so common/evergreen that, if a cheaper model made all the different language varients and ways of asking the question into one query, they could be cached quite effectively.

> 90+% of queries are probably so common/evergreen that, if a cheaper model made all the different language vari[a]nts and ways of asking the question into one query, they could be cached quite effectively.

As I wrote in some parallel post:

"When I look at my history with AI chatbots, 90 % of the conversations revolve around research-level problems in math and computer science. :-D"

For quite a few of them, creating a decent answer involved quite a bit of thinking by the AI, so I assume they are not the easiest queries to answer, and the queries are too obscure to cache.

Re: Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

#227

Post by khalic »

pixl97 wrote: post #222
khalic wrote: post #191

Earlier quoted context omitted.

That’s his definition and in no way a universal one. In the meantime, it’s still very easy to differentiate between an AI and a human in a chat. You just need to know the quirks of these systems. Like counting letters, hitting the safeguards, etc. So call me when one of them can pass the Turing test against me and then we can talk about AGI

There is no accepted definition of intelligence that is usable for classification of intelligence across the sciences. On top of that for your strong feelings, you don't have the conviction to write down a strong definition of intelligence yourself, which allows you to accelerate the goal posts up to light speed. The fun thing about writing out a formal definition is suddenly almost everything or almost nothing, incl…

[dead]

Re: Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

#228

Post by aleph_minus_one »

Tostino wrote: post #62
johnsmith1840 wrote: post #59

Earlier quoted context omitted.

I really don't see how they're going to be google here. The free tier is dominated by verticle integration and the platform that people use. Long term I imagine google wins the bottom of the market and I'd be suprised if they lost.

People tell these models everything . You don't think a pretty unethical company can figure out a way to monetize that?

> People tell these models everything. You don't think a pretty unethical company can figure out a way to monetize that?

I assume research-level questions about scientific topics (with a focus on math and computer science) are not easy to monetize (yes, this is by far the most common kind of question to AI models for me). :-D

Re: Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

#229

Post by throwaway473825 »

firasd wrote: post #7

I think it’s a misread to think the default ChatGPT model switching to GPT 5.6 Luna is some sort of desperation move. Keep in mind that Claude .ai never had this extreme stratification between the frontier models and the free tier (Sonnet is available to free users with rate limits). So 5.6 Luna is just their next version of what they used to call 5.5 instant tier And 5.x instant models were never much to write home…

I think the main competitor to OpenAI's free tier is Google Search's AI Overview powered by Gemini 3.5 Flash-Lite. OpenAI's free tier needs to be notably better than that for people to bother. There's also Gemini 3.6 Flash with a generous free tier.

Re: Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

#230

Post by kristofferR »

aleph_minus_one wrote: post #226
kristofferR wrote: post #85

Earlier quoted context omitted.

90+% of queries are probably so common/evergreen that, if a cheaper model made all the different language varients and ways of asking the question into one query, they could be cached quite effectively.

> 90+% of queries are probably so common/evergreen that, if a cheaper model made all the different language vari[a]nts and ways of asking the question into one query, they could be cached quite effectively. As I wrote in some parallel post: "When I look at my history with AI chatbots, 90 % of the conversations revolve around research-level problems in math and computer science. :-D" For quite a few of them, creating…

Sure, but you don't represent the average ChatGPT free user in that case probably. I regularly do super complicated queries that Pro spends ages replying to, but yesterday I also asked if chilis are toxic, or just unpleasant, for dogs, and if other species other than flamingos get colored by the food they eat.
Post reply on HN