381–390 of 445 posts
Earlier quoted context omitted.
Pixel hasn’t even been around for a decade.
The Pixel branding is 12 years old, and IIRC this feature also existed in Nexus before that.
Pixel phone launched in 2016.
Earlier quoted context omitted.
Doesn't make it untrue. Agents that can search the internet exist for a while now and have been essentially solved and happily used in platforms like Perplexity. It's really "meh", very far from revolutionary. Keep in mind this company is trying to convince everybody they need 500B USD now (through the Stargate project).
Let us know when your Bing-bot scores over 20% on the HLE benchmark.
Without internet: 10%
With internet: 23%
In addition:
> We found that the ground-truth answers for one dataset were widely leaked online
in very small letters, and they blocked these URLs at runtime but not training time.
It's not bad, but not revolutionary at all compared to the leap that was GPT-2 from GPT-3, or GPT-4o to DeepSeek-R1
I just gave it a whirl. Pretty neat, but definitely watch out for hallucinations. For instance, I asked it to compile a report on myself (vain, I know.) In this 500-word report (ok, I'm not that important, I guess), it made at least three errors. It stated that I had 47,000 reputation points on Stack Overflow -- quite a surprise to me, given my minimal activity on Stack Overflow over the years. I popped over to the l…
"Pretty neat, but definitely watch out for hallucinations." We'd never hire someone who just makes stuff up (or at least keep them employed for long). Why are we okay with calling "AI" tools like this anything other than curious research projects? Can't we just send LLMs back to the drawing board until they have some semblance of reliability?
Or is the position of OpenAI that Wiles' proof is incomplete?
For “deep research” I’m also reading “getting the answers right”. Most people I talk to are at the point now where getting completely incorrect answers 10% of the time — either obviously wrong from common sense, or because the answers are self contradictory — undermines a lot of trust in any kind of interaction. Other than double checking something you already know, language models aren’t large enough to actually kno…
I'd say they don't know anything.
An LLM base model, before it is post-trained with RL, just has access to a sliced and diced corpus of human output. Take the contents of 4chan and WikiPedia, put in blender and mix and chop into "training sample" sized bites, then learn the statistical regularities of this blended mess. It is what it is - not exactly what I'd call a knowledge base, even though there are bits of knowledge in there.
When you add RL-based post-training for reasoning, all you are doing is trying get the model to be more selective when you are sampling from it - encouraging it to suppress some statistics, and emphazise others, such that when you sample from it the output looks more like valid reasoning steps and/or conclusions, per the verified reasoning examples you train it on.
I'm well aware of how useful RL-tuned models (whatever the goal) can be, but at the end of the day all they are doing is taking a statistical babbler and saying "try to output patterns more like this". It's not exactly a recipe for factuality or rationality - we've just gone from hallucination-prone base models, to gaslighting-prone RL-tuned "reasoning" models that output stuff that sounds like coherent reasoning.
What missing from all of this - what makes it different from how animals learn - it that the model has no experience of it's own, no autonomy or motivation to explore, learn and verify, and hence no episodic memories of how it learnt something (tried it and ran controlled experiments, or just overheard it on the bus), and what that implies about it's trustworthiness.
It's amazing that LLMs work as well as they do, a reflection of how much of what we do can be accomplished by reactive pattern matching, but if you want to go beyond that to something that can learn and discern the truth for itself, this seems the wrong paradigm altogether.
Earlier quoted context omitted.
That is... debatable. You may be entirely inside the bubble, there.
Not sure if this was posted as humour, but I don't feel that way. In today's world, where I certainly would consider taking the blue pill, I'm having a blast with LLMs! It has helped me learn stuff incredibly faster. Especially I find them useful for filling the gaps of knowledge and exploring new topics in my own way and language, without needing to wait an answer from a human (that could also be wrong). Why does it…
In the early days of ChatGPT where it seemed like this fun new thing, I used it to "learn" C. I don't remember anything it told me, and none of the answers it gave me were anything that I couldn't find elsewhere in different forms - heck I could have flipped open Kernighan & Ritchie to the right page and got the answer.
I had a conversation with an AI/Bitcoin enthusiast recently. Maybe that already tells you everything you need to know about this person, but to the hammer the point home, they made a claim to similar to you: "I learn much more and much better with AI". They also said they "fact check" things it "tells" them. Some moments later they told me "Bitcoin has its roots in Occupy Wall Street".
A simple web search tells you that Bitcoin is conceived a full 2 years before Occupy. How can they be related?
It's a simple error that can be fact checked simply. It's a pretty innocuous falsity in this particular case - but how many more falsehoods have they collected? How do those falsehoods influence them on a day-by-day basis?
How many falsehoods influence you?
A very well meaning activist posted a "comprehensive" list of all the programs that were to be halted by the grants and loans freezes last week. Some of the entries on the list weren't real, or not related to the freeze. They revealed they used ChatGPT to help compile the list and then went down one-by-one to verify each one.
With such meticulous attention to detail, incorrect information still filtered through.
Are you sure you are learning?
Earlier quoted context omitted.
Anyone selling anything would want to remain crawlable if people use this to research something that could lead to a purchase.
Not necessarily. Southwest airlines doesnt allow itself on price comparison sites or Google Flights. Amazon listings are blocked from google shopping and other price comparison sites.
I see Amazon results there all the time. 3 of the visible 8 sponsored results are Amazon, in the non-sponsored results an Amazon listing is either first or second in every category.
Earlier quoted context omitted.
Let us know when your Bing-bot scores over 20% on the HLE benchmark.
It's literally a browsing agent that searches the internet and they know the questions in advance when preparing the agent Without internet: 10% With internet: 23% In addition: > We found that the ground-truth answers for one dataset were widely leaked online in very small letters, and they blocked these URLs at runtime but not training time. It's not bad, but not revolutionary at all compared to the leap that was GP…
Again: the assertion was yours, so let us know the results of your own work.
Earlier quoted context omitted.
>It has helped me learn stuff incredibly faster. Especially I find them useful for filling the gaps of knowledge and exploring new topics in my own way and language and then you verify every single fact it tells you via traditional methods by confirming them in human-written documents, right? Otherwise, how do you use the LLM for learning? If you don't know the answer to what you're asking, you can't tell if it's lyi…
I guess it all depends on the topic and levels of trust. How can I be certain that I have a brain? I just have to take something for granted, don't I? Of course I will "verify" the "important stuff", but what is important? How can I tell? Most of the time only thing I need is a pointer in the right direction. Wrong advice? I know when I get there I suppose. I can remember numerous things I was told while growing up,…
Looking in a resource written by someone with sufficient ethos that they can be considered trustworthy .
> What is real?
I'm not arguing ontology about systems that can't do arithmetic. you're not arguing in good faith at all
Earlier quoted context omitted.
Not sure if this was posted as humour, but I don't feel that way. In today's world, where I certainly would consider taking the blue pill, I'm having a blast with LLMs! It has helped me learn stuff incredibly faster. Especially I find them useful for filling the gaps of knowledge and exploring new topics in my own way and language, without needing to wait an answer from a human (that could also be wrong). Why does it…
Are you sure it's helped you learn? In the early days of ChatGPT where it seemed like this fun new thing, I used it to "learn" C. I don't remember anything it told me, and none of the answers it gave me were anything that I couldn't find elsewhere in different forms - heck I could have flipped open Kernighan & Ritchie to the right page and got the answer. I had a conversation with an AI/Bitcoin enthusiast recently. M…
Maybe we're already at AGI and just don't know it because we overestimate the capabilities of most humans.