dmitriy
@DmitriyLeybel
Followers
737
Following
51K
Media
1K
Statuses
7K
Building DuxMedic: an AI health platform to log symptoms, research conditions, explore treatments & guide your care to help you in your most stressful moments.
Oregon, USA
Joined January 2014
Prediction: Content provenance verification will be widespread extremely soon. Given the sanity-squishing AI slopification of the commons, it's not hard to predict that content provenance will soon be a first class citizen in social media. I realized from first principles that
1
0
6
When you open Microsoft Powertoys, it literally turns your entire PC theme to Light Mode. I am so done.
0
0
0
Me after Fogo de Chao
An enormously fat polar bear that has been gorging itself on the carcass of a whale on the coast of Norway.
0
0
0
The microplastics melted and are now nanoplastics
I reduced microplastics in my sperm by 85% and I think it was the dry sauna
0
0
0
https://t.co/ChH8TKRpcm Battle of the +.2 models
scoop: OpenAI’s GPT-5.2 "code red" response to Google is coming next week. I'm hearing that GPT-5.2 should drop on December 9th, slightly earlier than OpenAI was originally planning. Details here 👇
0
0
0
Money talks. Make sure you note the creators who release things with a lifespan of one year just to be bought out into giant slop machines and then publish scripted feigned-excitement statements like these.
I'm excited to share that Limitless has been acquired by Meta! Here’s why we joined forces, what this means for customers, and what comes next.
0
0
1
The way it handles the momentum is insanely impressive. Big if true
0
0
1
if I press play on my phone, I do not want spotify to play on my PC whose bright idea was this?
0
0
0
Modern LLMs/agents represent a step up the step function in our progress toward ASI. A very necessary step. What separates humans from other apes is our mastery of language and our relationship to information. The same holds true for endowing machines with a similar leap.
Social media tends to frame AI debate into two caricatures: (A) Skeptics who think LLMs are doomed and AI is a bunch of hype. (B) Fanatics who think we have all the ingredients and superintelligence is imminent. But if you read what leading researchers actually say (beyond the
0
0
0
How to be happy: Ignore everything he said and enjoy your Thanksgiving meal that comes once a year. Feasting has existed since the dawn of time.
Ave thanksgiving meal is = to smoking 7 cigarettes. Why and how to avoid... Average + 4,500 calories across day + 229 g fat + 69 g saturated fat + 450–600 g carbs + 150–200 g sugar (pies, rolls, alcohol...) + 3,000–4,500 mg sodium The damage: + massive glucose spikes +
0
0
2
Don't mix AI with Adderall. Your slop will grow slop cancer.
"Izugaku Never Say Goodbye" may go down as one of the best AI songs of all time. It was created by a YouTuber named Steve Terreberry who typed a bunch of random letters into a song generator and hit "optimize." Somehow, the result was a total banger.
0
0
0
If Linux could reliably run videogames, I wouldn't touch a MS product ever again in my life.
0
0
0
I wonder which of our benevolent bureaucrats are getting lifetime advisory positions and mansions in a few years....
The EPA is starting to allow the use of pesticides containing PFAS on food. The move is part of an effort to roll back the regulation of what are called "forever chemicals" because they don't break down easily in the environment.
0
0
0
Critique: Should be 5 hours
The @ilyasut episode 0:00:00 – Explaining model jaggedness 0:09:39 - Emotions and value functions 0:18:49 – What are we scaling? 0:25:13 – Why humans generalize better than models 0:35:45 – Straight-shotting superintelligence 0:46:47 – SSI’s model will learn from deployment
0
0
0
One of the easiest ways to digest information coming from an LLM is to use "Create a mermaid chart of this..." I find that this gives a great overview most of the time.
0
0
0
Same. How would we even begin to address this? We have the ability to verify other digital media through a set of cryptographic signatures associated with the source and edited content, but we can't really do that with the sheer amount of AI text slop. Even when verbalized,
I’m actually being driven crazy by this Otherwise smart mutuals on Twitter commenting on AI-written slop essays Family members reposting obviously-AI generated text on Facebook Apparently the NYT editorial staff can’t spot it I’m moving to a cabin in the woods
0
0
0