AI Post β€” Artificial Intelligence
657K subscribers
3.99K photos
2.9K videos
3 files
5.35K links
πŸ€– The #1 AI news source! We cover the latest artificial intelligence breakthroughs and emerging trends.

Manager: @rational
Download Telegram
Media is too big
VIEW IN TELEGRAM
🧬 AI just built a world model of the human cell

GenBio AI has previewed AIDO Cell, a new AI system designed to simulate a living cell across DNA, RNA, proteins, regulatory networks, and whole-cell behavior, all inside one model.

Instead of having separate models predict what happens at each biological layer, AIDO Cell connects the entire chain. Change a gene, overexpress one, or introduce a compound, and the effects can propagate through the cell from gene regulation to RNA, proteins, compartments, and ultimately cell behavior.

AIDO Cell remembers the state of the cell. Make a change, then make another one and the second intervention acts on the cell you just changed, not a fresh copy of the original.

That turns it from something you simply query into something you can actually experiment on.

It’s still an early preview, not a perfect digital replica of any specific cell. But if this approach scales, AI could eventually let researchers run complex biological experiments inside a virtual cell before touching a real one.

Source.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ‘258🀑257πŸ‘€247❀217
Media is too big
VIEW IN TELEGRAM
Me explaining AI to friends

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
😁248πŸ€”229πŸ—Ώ204❀176πŸ‘176
πŸ‘€ DeepSeek just gave AI agents eyes

DeepSeek has released DeepSeek-V4-Flash-Vision-Exp, an experimental multimodal model that brings vision to its V4-Flash capabilities.

The model reportedly matches V4-Flash on text tasks like reasoning, agents, and world knowledge, while making a big jump on multimodal agent benchmarks, pushing performance close to Opus-4.8.

That means AI agents can now understand images while using tools, opening up workflows that previously required separate vision models.

Source.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❀288πŸ‘Ύ261πŸ‘227πŸ‘€204
❗️OpenAI and Anthropic just folded on their data retention policy after Palantir CEO exposed their real business model

Alex Karp:

"something has gone completely wrong"
they want "access to my data" so they can "build my alpha"
"i'm gonna get no value, and they're gonna get my IP"
"this is insane"
"trying to drug addict us to a future they believe they control"
"who owns the data? where is it cached? are the prompts secured?"
"we need to rebuild trust"

Yesterday: OpenAI started testing "private safety processing" to avoid retaining customer data

Today: Anthropic reversing course, will allow enterprise customers to keep data on THEIR OWN cloud infrastructure instead of Anthropic's

Alex Karp was right.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ‘Ύ207🫑200😐193πŸ‘178❀165
This media is not supported in your browser
VIEW IN TELEGRAM
Don't worry, the robots won't be coming for your jobs just yet.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
😁235πŸ‘200🍌180πŸ‘Ύ177πŸ‘€175
NVIDIA’s AI coding agent AVO has achieved a 100% score on the ARC-AGI-3 interactive reasoning benchmark, solving all 183 levels across 25 public environments.

AVO operates without predefined rules or explicit goals. It learns by experimenting, observing outcomes, and adjusting its strategy based on past experiences. Unlike previous models, AVO maintains and builds on its learning through context resets rather than starting from scratch.

The system is powered by Claude Opus 5, which has previously been used to autonomously optimize GPU code over several days.

This result demonstrates AVO’s ability to master environment-specific tasks through repeated interaction and continuous memory updates.

πŸ“° @aipost
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ‘324❀297πŸ™287πŸ‘€16
Media is too big
VIEW IN TELEGRAM
At World Robot Conference, Elf-Xuan 2.0 from AheadForm delivered a live vocal performance with real-time lip and expression sync plus arm gestures.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ‘264🍌240❀228πŸ‘Ύ212
❗️OpenAI VP of Sales just resigned after only 5 months… a week after her boss, the Chief Revenue Officer also quit.

β€œMany other people on the sales team are considering resigning”

The exodus continues.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ€ͺ279❀216πŸ‘183🀑177πŸ€”90
⚑️AI RACE: OpenAI and Anthropic are in an all-out war.

OpenAI is cutting GPT-5.6 Sol API prices by over 20% for the next 3 months. It launched ads inside ChatGPT this month and rolled out 14x faster inference.

Meanwhile Anthropic just hit $65 BILLION in annual revenue and is building its own custom chips. It is targeting a $2 TRILLION IPO in October while OpenAI's has been delayed to 2027.

Anthropic leads U.S. enterprise adoption 43.5% to 39.7%, but OpenAI has 900 MILLION weekly users to Anthropic's 245 MILLION monthly, per Ramp.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
❀324πŸ‘299πŸ”₯298
πŸ“’ Behind every good AI answer is a person who graded a worse one.

Here's how to use that:

Before a model like ChatGPT or Claude reaches you, thousands of people sit in front of its responses and grade them. They compare two answers and mark which one is better, flag replies that are wrong or unhelpful, and rewrite weak responses into stronger ones. The model studies those judgments and shifts toward the kinds of answers people actually rate as good.

What the model picks up from all that grading is a sense of what a useful answer looks like to a real person. That's the quiet reason a chatbot can feel helpful rather than merely correct.

Here's the part you can use, because the same loop works inside a single chat. When an answer misses, don't just rephrase and hope for better. Tell it what was wrong, show it what a strong version would include, and ask it to revise against that. Rate it the way a grader would, and it adjusts on the spot.

The more specific your feedback, the closer the next answer lands.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ—Ώ307❀284πŸ‘277
Media is too big
VIEW IN TELEGRAM
A robot guide dog successfully led a journalist through a simulated busy airport terminal at the 2026 World Robot Conference in Beijing.

The demonstration showed the robot assisting with navigation to a check-in counter and finding the correct gate. The technology has the potential to offer practical assistance for seniors if made widely available.

The demonstration was part of ongoing efforts to develop affordable and accessible robotic support systems.

πŸ“° @aipost
Please open Telegram to view this post
VIEW IN TELEGRAM
❀209πŸ”₯187πŸ‘176πŸ™170πŸ‘Ύ170
UBS projects $4.1 trillion in AI infrastructure spending between 2026 and 2028, exceeding three times the total outlay of the last six years.

The top spenders include Google, Meta, Microsoft, and Amazon, with projected three-year investments of $938 billion, $683 billion, $672 billion, and $628 billion respectively. Neocloud and China Cloud providers such as CoreWeave and Nebius are also rapidly increasing their share.

Nvidia and Broadcom are identified as primary beneficiaries, supplying AI accelerators and advanced networking components. Power equipment manufacturers including Eaton, Vertiv, and GE Vernova have seen substantial gains, responding to growing demand for transformers and cooling systems.

Memory companies Samsung, SK Hynix, and Micron have sold out their 2026 capacity for high bandwidth memory. ASML and TSMC are crucial to advanced chip production needed for this AI expansion.

πŸ“° @aipost
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ‘€270πŸ₯΄223❀206πŸ€”175πŸ‘73
This media is not supported in your browser
VIEW IN TELEGRAM
A new Chinese rescue robot is now operating autonomously to assist drowning individuals. It deploys by air, flying to the location at speeds of approximately 10–14 meters per second and can travel up to 2 kilometers.

Upon reaching the person in distress, the robot lands on the water to provide flotation, supporting up to two adults with an approximate weight of 80 kilograms each. After the rescue is complete, the robot is capable of returning independently.

πŸ“° @aipost
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ‘386❀363🀑234
πŸ’° $65 BILLION revenue run rate.

Anthropic could theoretically be growing insanely fast AND losing share at the same time.

Which basically means: THE WHOLE MARKET IS GROWING EVEN FASTER.

At these numbers and potential trillion-dollar valuations...

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❀272πŸ‘Ύ242πŸ‘220🀑175😐47
This media is not supported in your browser
VIEW IN TELEGRAM
Incredible AI robots are being built in china.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
😁391❀236🀑186πŸ‘79πŸ”₯77