Richard Davies

Founder @ Meridias | Artificial Intelligence Expert | Generative AI | Large Language Models | Agents | NLP | Computer Vision | Published Author

[Richard Davies @RichardDaviesAI]

https://t.co/KK6NoFez9f

[Richard Davies @RichardDaviesAI]

@gabriberton I always wondered why the term "meat proxy" came about. This post makes it fairly clear.

Rather than explain why Chollet is wrong, you've asked several LLMs and treated their agreement as proof. Present an argument, otherwise you're just adding noise.

[Richard Davies @RichardDaviesAI]

@petergyang The number of people in this thread who haven't heard of symlinks is crazy.

[Richard Davies @RichardDaviesAI]

One of the most frustrating phrases for Domain Experts is quickly becoming:

"Well, ChatGPT said this..."

Followed immediately by someone with little to no Domain Knowledge proceeding to lecture others while speaking absolute nonsense with complete confidence.

[Richard Davies @RichardDaviesAI]

@pmarca You've said for weeks that jobs won't be replaced, and now you’re claiming we've achieved AGI.

AGI means general intelligence across domains. If so, displacement follows.

So how do you reconcile that with jobs not being replaced?

[Richard Davies @RichardDaviesAI]

@SchmidhuberAI Would you consider publishing a book in the future to correct the record?

• Talking Nets: An Oral History of Neural Networks
• The Quest for Artificial Intelligence

And a few others, tend to over credit Hinton and LeCun, with little mention of Rumelhart, Werbos, and others.

[Richard Davies @RichardDaviesAI]

@atuntable @kareem_carr Invoking "natural selection" is a category error, it borrows a biological metaphor to explain a system that is explicitly designed, trained, and deployed by humans.

I'm going to step away from this discussion. Enjoy your day.

[Richard Davies @RichardDaviesAI]

@atuntable @kareem_carr The "house on fire" framing isn't grounded in how Transformer based Language Models, or AI techniques more broadly, actually behave. It is entirely driven by investor optics and commercial incentives.

Without technical grounding, there isn't anything substantive to discuss.

[Richard Davies @RichardDaviesAI]

@craigzLiszt This is classic AI grifter rhetoric, provocation dressed up as insight.

Employability has no intrinsic relevance to relationships. Its importance arises only within transactional models, not within stable, long term partnerships.

[Richard Davies @RichardDaviesAI]

@haider1 I agree with @_KarenHao. "AGI" is optics driven messaging targeted at investors who lack meaningful domain understanding of Artificial Intelligence.

[Richard Davies @RichardDaviesAI]

@asmah2107 I agree. From the business side it is largely a rebrand for optics, while for employees it is bandwagon driven hype tied to salary expectations. Software Engineers have long used AI based techniques, including pre-trained Machine Learning models.

[Richard Davies @RichardDaviesAI]

@zaph0id @ericw_ai Yes

[Richard Davies @RichardDaviesAI]

@elonmusk I'd say it would be remarkable to have a small village of 16 permanently based on Mars in 30 years.

One million people seems like an overestimate for that timeframe.
However, I wish you the best of luck with it.

It’s a bold vision, and bold visions change the world.

[Richard Davies @RichardDaviesAI]

@Grady_Booch @netterrain @IBMResearch I hope you enjoy your day @Grady_Booch, I live near the sea but unfortunately there's no whales here.

[Richard Davies @RichardDaviesAI]

@Grady_Booch I bet your library is a treasure trove of knowledge. I've been building mine for the past seven years. I have roughly 500 to 600 books on Artificial Intelligence and recently started reading on Software Architecture, as inspired by you!

[Richard Davies @RichardDaviesAI]

The problem with calling Chain of Thought reasoning (whether through tokens or the latent space representation) is that any step by step instructions in the training data (whether pre-training, fine-tuning, or RFHL) are merely representations of the solution derived from an earlier decision making process. All options are considered, decisions made, and intentionality occurs prior to putting pen to paper, this is true of source code too.

Even if one were to write down the entire decision making process, including all options considered and the intention behind the decision, it would still lack intrinsic meaning. Language is merely a communication tool. Humans recognise patterns in the data (natural language) and interpret them into internal abstract representations. The meanings of symbols themselves are not stored within the data, a concept well known in philosophical circles as the Symbol Grounding Problem.

Acknowledging this makes it clear that Language Models do not reason but rather reproduce patterns created by humans. This ability to reproduce patterns still remains extremely useful for automation.

#artificialintelligence #data #philosophy

[Richard Davies @RichardDaviesAI]

@fchollet I wonder how the previous generation of leaders in Artificial Intelligence, such as McCarthy, Schank, and Lenat, would view non-scientific or subjective evaluation methods such as "vibes". It seems unlikely they would endorse it. I guess this is a sign of the times.

[Richard Davies @RichardDaviesAI]

@joecarlsonshow Meta AI is a leader in the Artificial Intelligence, headed by Yann LeCun, one of the most prominent technical minds in field.

They have been advocating for open source language models. As well as developing the LLAMA series, which is widely used across the technology industry.

[Richard Davies @RichardDaviesAI]

Here's a link to my Article:

Why Fully Autonomous Software Engineering with AI Remains a Distant Dream

https://t.co/iDfUY4mGLo

@GaryMarcus @dmonett @Grady_Booch

[Chubby♨️ @kimmonismus]

In 2025: code like mid-level engineers. Eventually, AI engineers will build most of the code and AI in apps, replacing human engineers. You heard it directly from Zuck. AI will replace your job. No denying anymore.