Ricardo Leclaire

Psychodynamic psychotherapist, science and tech enthusiast, with specific interest in AI. Also some background in software engineering and biomedical sciences.

[Ricardo Leclaire @Ric_luchtlek]

@0xJokker The AMD Ryzen Max+ 395 has been around for.... like a year?

[Ricardo Leclaire @Ric_luchtlek]

@UserJourneys @sama Agreed, although custom instructions do help here.

[Ricardo Leclaire @Ric_luchtlek]

@Sethmr @sairahul1 More and more people can easily run models like this locally. With a Macbook with anything over 16GB ram it's no issue. And a lot of people doing this kind of thing have more powerful machines. You don't, that's clear, but that isn't saying much.

[Ricardo Leclaire @Ric_luchtlek]

@flamz321 @sairahul1 What are you talking about? This runs locally, none of it goes to Microsoft's servers...

[Ricardo Leclaire @Ric_luchtlek]

@VraserX It really won't....

[Ricardo Leclaire @Ric_luchtlek]

@cryptopunk7213 Yes, it's a newer model. Of course it did. Also 4.7 was just an incremental update, 5.5 was marketed as an entirely new base model, hyped up, and in comparison to the hype it was (once again) really disappointing if you ask me.

[Ricardo Leclaire @Ric_luchtlek]

@S86067S @kimmonismus How much work do they do without a prompt though? And how much can they actually learn and retain? And how is their world model? What they're doing is essentially prompting themselves into next steps, or you could say, answering their own questions.

[Ricardo Leclaire @Ric_luchtlek]

@AluminaGu @S86067S @kimmonismus No, this guy actually isn't. This guy is subscribed to a ChatGPT Plus account and Claude Max 20x account using Claude Code daily.

[Ricardo Leclaire @Ric_luchtlek]

@haider1 I’m curious which AI company will be the first to dare to really investigate and invest in new architectures, specifically dynamically updating self learning models, ideally with world models.

[Ricardo Leclaire @Ric_luchtlek]

@0xRizzler @kimmonismus It's a boring iteration is what it is. What we "need" is a new architecture and an actual step forward into something new, like models that dynamically update themselves.

[Ricardo Leclaire @Ric_luchtlek]

@musiol_martin @kimmonismus How sure are you about that? Because AFAIK that's what they've been doing since o1. My understanding was that this was different, although I don't expect anything worthy of the hype it's given.

[Ricardo Leclaire @Ric_luchtlek]

@gdb What about clinicians outside the US?

[Ricardo Leclaire @Ric_luchtlek]

@CodyAlanAllred @VraserX Agree

[Ricardo Leclaire @Ric_luchtlek]

@VraserX It’s sort of a new pretrain (they’ve been vague about this and said that it’s akin to it). That doesn’t mean it’s significantly beyond anything we’ve seen and I suspect disappointment. It will be a bit better and that’s it.

[Ricardo Leclaire @Ric_luchtlek]

@MilkRoadAI Accident. Of course. Totally woopsie

[De Speld @DeSpeld]

Dear President @realdonaldtrump this is what our failing right leader @GeertWilders is doing, desperately seeking attention ratting out our prime minister and perfectly willing to hurt our national interests. Please don’t trust him, he is the biggest loser of the last elections, a fake nationalist like yourself, with no friends left. Sad and pathetic! Thank for your attention to this matter.

[Ricardo Leclaire @Ric_luchtlek]

@kimmonismus He saw dollars

[Ricardo Leclaire @Ric_luchtlek]

@sukh_saroy Whether this is true or not is rather irrelevant imo. It's similar to saying 20 years ago that kids these days can't do simple calculations off of the top of their heads anymore. Maybe it's just not that relevant.

[Ricardo Leclaire @Ric_luchtlek]

@HelicalDee @DeryaTR_ Their timeline also shortened... from 2030 to 2028...