The popular online discourse on Large Language Models’ (LLMs’) capabilities is often polarized in a way I find annoying and tiresome.

On one end of the spectrum, there is nearly complete dismissal of LLMs: an LLM is just a slightly fancier version of the autocomplete on your phone’s keyboard, there’s nothing to see here, move on (dot org).

This dismissive perspective overlooks some genuinely interesting novel capabilities of LLMs. For example, I can come up with a new joke and ask ChatGPT to explain why it’s funny or come up with a new reasoning problem and ask ChatGPT to solve it. My phone’s keyboard can’t do that.

On the other end of the spectrum, there are eschatological predictions: human-level or superhuman artificial general intelligence (AGI) will likely be developed within 10 years or even within 5 years, and skepticism toward such predictions is “AI denialism”, analogous to climate change denial. Just listen to the experts!

There are inconvenient facts for this narrative, such as that the majority of AI experts give much more conservative timelines for AGI when asked in surveys and disagree with the idea that scaling up LLMs could lead to AGI.

The ARC Prize is an attempt by prominent AI researcher François Chollet (with help from Mike Knoop, who apparently does AI stuff at Zapier) to introduce some scientific rigour into the conversation. There is a monetary prize for open source AI systems that can perform well on a benchmark called ARC-AGI-2, which recently superseded the ARC-AGI benchmark. (“ARC” stands for “Abstract and Reasoning Corpus”.)

ARC-AGI-2 is not a test of whether an AI is an AGI or not. It’s intended to test whether AI systems are making incremental progress toward AGI. The tasks the AI is asked to complete are colour-coded visual puzzles like you might find in a tricky puzzle game. (Example.) The intention is to design tasks that are easy for humans to solve and hard for AI to solve.

The current frontier AI models score less than 5% on ARC-AGI-2. Humans score 60% on average and 100% of tasks have been solved by at least two humans in two attempts or less.

For me, this helps the conversation about AI capabilities because it gives a rigorous test and quantitative measure to my casual, subjective observations that LLMs routinely fail at tasks that are easy for humans.

François Chollet was impressed when OpenAI’s o3 model scored 75.7% on ARC-AGI (the older version of the benchmark). He emphasizes the concept of “fluid intelligence”, which he seems to define as the ability to adapt to new situations and solve novel problems. Chollet thinks that o3 is the first AI system to demonstrate fluid intelligence, although it’s still a low level of fluid intelligence. (o3 also required thousands of dollars’ worth of computation to achieve this result.)

This is the sort of distinction that can’t be teased out by the polarized popular discourse. It’s the sort of nuanced analysis I’ve been seeking out, but which has been drowned out by extreme positions on LLMs that ignore inconvenient facts.

I would like to see more benchmarks that try to do what AGI-AGI-2 does: find problems that humans can easily solve and frontier AI models can’t solve. These sort of benchmarks can help us measure AGI progress much more usefully than the typical benchmarks, which play to LLMs’ strengths (e.g. massive-scale memorization) and don’t challenge them on their weaknesses (e.g. reasoning).

I long to see AGI within my lifetime. But the super short timeframes given by some people in the AI industry feel to me like they border on mania or psychosis. The discussion is unrigorous, with people pulling numbers out of thin air based on gut feeling.

It’s clear that there are many things humans are good at doing that AI can’t do at all (where the humans vs. AI success rate is ~100% vs. ~0%). It serves no constructive purpose to ignore this truth and it may serve AI research to develop rigorous benchmarks around it.

Such benchmarks will at least improve the quality of discussion around AI capabilities, insofar as people pay attention to them.

I’ve been wondering about this lately.

Recently, I made a post about Ukraine on another social media site, and someone jumped in with the usual "Ukraine isn't a democracy" right-wing talking point. I wrote out a long, thoughtful reply, only to get the predictable one-liner propaganda responses back. You probably know the type, just regurgitated stuff with no real engagement.

After that, I didn’t really feel like spending my time and energy writing out detailed replies to every canned response. But I also didn’t want to just let it sit there and have people who might be reading the exchange assume there’s no pushback or correction.

So instead, I tried leveraging AI to help me write a fact-checking reply. Not for the person I was arguing with, really, but more as an FYI for anyone else following along. I made sure it stayed factual and based in reality, avoided name-calling, and kept the tone above the usual mudslinging. And of course, I double-checked what it wrote to make sure it matched my understanding and wasn’t just spitting out garbage or hallucinations.

But it got me thinking that there’s a lot of fear about AI being used to spread and create misinformation. But do you think there’s also an opportunity to use it as a tool to counter misinformation, without burning ourselves out in the process?

Curious how others see it.

I asked Claude for references to some ideas the ai had output earlier about a 16th century cardinal (Sadoleto). It's response was, Let me analyze the letter, and then this:

async function searchForSadoletoLetter() {
  try {
    const files = ['sadoleto_letter.txt', 'sadoleto.txt', 'sadoleto_to_geneva.txt', 'letter_to_geneva.txt'];
    
    for (const filename of files) {
      try {
        const content = await window.fs.readFile(filename, { encoding: 'utf8' });
        console.log(`Found file: ${filename}`);
        console.log("First 500 characters:");
        console.log(content.substring(0, 500));
        return content;
      } catch (err) {
        // File not found, try the next one
      }
    }
    console.log("Could not find any files containing Sadoleto's letter");
    return null;
  } catch (err) {
    console.error("Error searching for files:", err);
    return null;
  }
}

const content = await searchForSadoletoLetter();

// Since we don't have the letter available as a file, let's search for published versions online
console.log("The letter isn't available as a file. We would need to rely on scholarly knowledge about Sadoleto's letter to Geneva (1539).");
console.log("Sadoleto's letter was published in 'A Reformation Debate: Sadoleto's Letter to the Genevans and Calvin's Reply' (ed. John C. Olin).");

This strikes me as odd? Did Claude write a script to generate its own response? Have coders put something like this in as a gaurdrail?

edit: details about earlier convo

I recently had an unpleasant experience. Something I wrote fully and without AI generation of any kind was perceived, and accused of, having been produced by AI. Because I wanted to get everything right, in that circumstance, I wrote in my "cold and precise" mode, which admittedly can sound robotic. However, my writing was pointed, perhaps even a little hostile, with a clear point of view. Not the kind of text AI generally produces. After the experience, I started to think of ways to write less like an AI -- which, paradoxically, means forcing my very organic self into adopting "human-like" language I don't necessarily care for. That made me think that AI is probably changing the way a lot of people write, perhaps in subtle ways. Have you noticed this happening with you or those around you?

Every company is trying to shoehorn AI into every product, and many online materials provide a general snake oil vibe, making it increasingly difficult to parse. So far, my primary sources have been GitHub, Medium, and some YouTube.

My goal is to better understand the underlying technology so that I can manipulate it better, train models, and use it most effectively. This goes beyond just experimenting with prompts and trying to overcome guardrails. It includes running local, like Ollama on my M1 Max, which I'm not opposed to.

I'm working on an AI chat portal for teams, think Perplexity but trained on a company's knowledgebase (prosgpt dot com for the curious) and i wanted to talk to some people who are successfully using LLMs in their teams or jobs to improve productivity

Are you using free or paid LLMs? Which ones?

What kind of tasks do you get an LLM to do for you?

What is the workflow for accomplishing those tasks?

Cheers,
nmn

Has anyone out there hooked HomeAssistant up to a local LLM? I'm very tempted:

• Alexa integrations fail often. HomeAssistant integrations tend to be rock solid.
• Alexa is rule/pattern matching based. LLMs can understand natural language fairly well. The "magical incantations" required by Alexa are awkward.

Other than the software, the device side seems challenging. There are $50 fully-baked POP devices. I'm less sure on the DIY front.

Also, I desperately want my house to speak to me in the voice of the NCC-1701D computer. I've read enough now to know this should be achievable with a modicum of effort via OSS voice cloning tools or training a new model (same difference except "voice cloning" seems to often refer to doing this without training a whole new model?).

Thoughts? Experiences?

I've seen several pages that have led me to conclude this is tenable:

https://github.com/myshell-ai/OpenVoice

https://github.com/domesticatedviking/TextyMcSpeechy

https://github.com/mezbaul-h/june

https://www.home-assistant.io/voice_control/voice_remote_local_assistant/

https://heywillow.io/hardware/#esp32-s3-box-lite