Richard Socher

@RichardSocher124k followers

Building self-improving superintelligence CEO @recursive_si and @youdotcom MP @aixventuresHQ Ex: Stanford Adj Prof, Chief Scientist at Salesforce, CEO MetaMind

Open on X
Top outlier
155×
Best organic post
Peak views
2.7M
Single post
Bookmarks
13k
Across 95 organic posts
Last scan
Oct 16
Community cache
Posts 95 posts
#DatePost↓ OutlierViewsLikesRepostsBookmarks
1Dec 23, '22Today, youChat goes live. Open, broadly capable, conversational AI for search with knowledge of recent events and citations of sources. Search and chat of the future:155×2.7M4.5k7341.5k
2Jan 23, '23If you like #chatGPT -like capabilities and want to use conversational AI as your default search engine, try - it is free, has access to recent sources and often gives citations. Y58.9×1M1.2k97583
3Jun 21, '23Introducing You⚡Pro One AI chatbot to search the web, chat, code, write and create images. Now supercharged with GPT-4 and Stable Diffusion XL. Stay up-to-date with the newest AI m32.4×564k1.9k8685
4Mar 18, '23With this much excitement and energy in the LLM space, I am fairly confident in this prediction: Before the end of this year, there will be an open source model that is equivalent 28.8×502k1.1k119102
5Sep 20, '23Today, we launch YouAgent, an AI agent with code execution for more accurate answers to complex math and science questions. Unlike other consumer chat engines that cannot do reliab26.9×468k2.2k175394
6Feb 6, '23Do you like prompt engineering or sharing amazing #chatGPT - like transcripts with friends? We just launched sharing of full transcripts on youChat. The best part - you can continu23.9×416k1.1k3939
7Feb 19, '25It's always amazing to see your past interns and students lead multi-billion dollar companies. He's one of at least 4 of mine that I can think of. HuggingFace and Commure are two o20.8×363k3.4k103696
8Nov 23, '23Proposed Theorem: it’s impossible to acquire super humanity skills when relying purely on data created by humanity. Why do I say super humanity and not super human? Because for exa19.8×344k33131142
9Oct 24, '23Bored by generic chatbot answers? Today, @YouSearchEngine is moving closer to our vision of being your personal AI assistant for search. Introducing Smart Personalization ✨ No need18.3×319k2.2k118109
10Aug 30, '23Excited to introduce YouPro for Education—your AI study buddy. Access unlimited AI chat + search, unlimited AI writing generations, unlimited AI art generations, supercharged with 16.7×291k1.4k5496
11Sep 24, '23I struggle with this sometimes when early career folks ask me for advice and how they could be successful ... I ruined my health for a good decade working crazy hours but feel bad 16×279k39526161
12Apr 5, '23Hi @elonmusk, I would believe your concerns more if Tesla didn't download powerful AI into every car. Why over-regulate some AI models instead of real applications? Language models15.4×268k4933918
13Aug 5, '24The problems in a domain that you can simulate, can be solved by AI since you can create unlimited training data. This is true for games (think go, dota, etc), simple physical simu15×260k54273160
14Aug 22, '24If each step of an ai agent is 95% accurate. None of the 30 step work flows will work. Going from 95-> 99.9 is a similar last mile problem as with self driving cars. Easy to hac14.8×258k1.2k104288
15May 8, '23I think one of the problems of academic publishing/peer reviews is that there is zero downside of rejecting good ideas. The reviewers of the DecaNLP paper - which introduced prompt13.6×237k63894115
16Aug 7, '23Growing up in Germany you're taught that nuclear energy is terrible. Yet, all my scientific friends agree that it is by far the cleanest form of energy - it is SO much better for t12.1×211k7304325
17Apr 13, '23and this is why the front seats are not the best place to be in the worst case scenario :||12×210k3663821
18Jul 24, '24Now that the world has access to a GPT4 level model completely open source, we will see that the fear mongering AI p(doom)ers were wrong again about the supposedly existential risk11.5×201k6997250
19Mar 28, '23I think the open source community needs to start incorporating the concepts of * open source training data * open source human feedback collection * open source weights With LLMs i11.4×199k6098347
20Sep 17, '23I haven't been this excited about an announcement since we launched youChat last year (when we were the first to connect an LLM to the internet). LLMs have enabled new ways of lear11×191k3092187
21May 3, '22Introducing YouCode: The best search engine for developers. Available now at . A 🧵 with our features that make you more efficient when coding. We’d love your feedback to make it b10.8×02.6k442677
22Apr 29, '20Excited to introduce the AI Economist: Extends ideas from Reinforcement Learning for tackling inequality through learned tax policy design. The framework optimizes productivity and9.2×02k648314
23Apr 17, '23It is fascinating how the @nytimes can report on Google MAYBE adding "novel" features IN THE FUTURE without mentioning that @YouSearchEngine launched several of these features publ9.1×158k4374630
24Apr 19, '23Here's my mental model for when a generative ai application makes sense: GenAI is amazing when it would take a long time to create an artifact but very little time to verify its co8.8×153k46245179
25Dec 15, '22Today, we are announcing the opening of our search platform. With this, we hope to tackle two problems we see with Google: 1. users do not have control over what they see, and 2. d8.7×152k34241115
26Jun 22, '25If you studied algorithms, I'm sure you've heard of Dijkstra’s algorithm to find the shortest paths between nodes in a weighted graph. Super useful in scenarios such as road networ8.1×141k1.2k120918
27Mar 6, '24you vs pplx substance vs hype7.9×137k45626220
28Nov 20, '23The power of staying in one company, learning, improving, creating. A lot of young employees seem to think there's a need to jump around companies every few years and while you can7.9×137k39435132
29Feb 3, '25Now that openai has caught up to our old research agent from over a year ago we will soon launch our advanced research and intelligence agent which will be 10x better than what we 7.2×125k76050198
30Feb 27, '25We’re excited to introduce ARI (Advanced Research & Insights) - the first professional-grade deep research agent purpose-built for business. Instead of spending $100K+ on whitepape6.9×121k73470703
31Sep 11, '19We release the largest publicly available language model: CTRL has 1.6B parameters and can be guided by control codes for style, content, and task-specific behavior. Incredible gen6.9×01.6k566207
32Feb 8, '25Happy to dive into more details here but the main reasons why LLMs are not making amazing discoveries are: 1) passive vs active knowledge. Like when you can understand a language b6.9×120k58888476
33Apr 20, '23Humanity just unlocked nano-surgery: we can now inject magnetic carbon nanotubes into a (mouse) brain. They're engineered to ONLY bind to cancer cells. Then you turn on a magnetic 6.3×109k4456191
34Mar 12, '23This Monday I had the opportunity to advice the German chancellor and his cabinet on AI, its countless opportunities as well as risks and intuitions for GPT-like models.6.2×107k5783033
35Mar 25, '24This kind of blows my mind: Imagine YOU are a data scientist, get a data file and need to build a classifier... Just tell your AI assistant to "classify this" and ... it's all DONE5.9×103k46541397
36Oct 20, '23Does the graphics community have any ethics folks worrying about disinformation or is that more of an AI community thing? (Honest question, because I never hear about computer grap5.7×100k4165049
37Jun 20, '18Very excited to announce the natural language decathlon benchmark and the first single joint deep learning model to do well on ten different nlp tasks including question answering,5.7×01.3k53285
38May 13Launching Recursive Today, I’m very excited to announce the launch of @Recursive_SI, our new company building recursive self-improving superintelligence that automates knowledge di5.2×91k46157244
39Oct 25, '23We are now offering the first full web index API for LLMs. It already works better than Bing & Google SERP APIs for making LLMs more factual, up-to-date and offer citations. It4.7×82k90179145
41Dec 16, '23I'm happy for Tomas Mikolov and his co-authors to have won the test of time award. It's well deserved and I congratulated him on Facebook already a few days ago. He is Schmidhuber-4.5×78k35434100
42Jan 23, '23YouImagine now has a MidJourney option and boy - it's faster AND creates a broader range of styles: photorealistic/paintings/drawings/etc. #ai #generativeart #GenerativeAI4.5×78k3042776
43May 5, '22Presenting: The AI Economist This is one of the most impactful lines of AI research I've ever worked on. Its implications span from immediately impactful to highly philosophical. B4.2×0794149391
44Dec 14, '23Some predictions for AI in 2024: We're going to see open source models catch up to GPT4 and companies eventually using LLM operating systems that help you get these models producti4.2×72k27848108
45Jul 13, '20I will be starting a new company -- with a vision I have been passionate about for a long time.4.1×01.6k5617
46Nov 14, '23Want your chatbot to be up-to-date? Our new YOU API empowers your LLM with real-time web search, much longer snippets and better fact recall than any other search API. We also have4.0×69k1.1k7539
47Apr 29, '22View post on X3.9×01.7k3925
48Jan 12, '24This year companies will realize that an LLM by itself, even when fine tuned, will not be accurate or up-to-date and know about the world in a verifiable way. That's why we're exci3.8×66k3022347
49Dec 7, '23With Google's Gemini coming out, there's pressure for every LLM to have a live web connection. Ping us at api@you.com if you want to get rid of hallucinations, keep your LLM answer3.8×66k6985114
50Jan 29, '24Wow. This is huge news for YOU. will be a default browser option for EU countries on iOS 17.4 We're the only chat-native browser option that made the cut. Read more:3.8×65k5125143
51Sep 3, '25We’re officially a YOUnicorn! 🦄 Excited to share that @youdotcom just raised $100M Series C at a $1.5B valuation, led by @CoxEnterprises. We’ve been heads down building the search3.7×64k4303092
53Dec 8, '21It is kind of shocking how much current computing architectures (graphics processing units) define neural network architectures. GPUs made models like Transformers fast and hence p3.4×01.1k12583
54Sep 4, '2425 years ago, knowing "how to google" gave you an edge when it came to being more productive. Today, it's about prompt engineering and creating AI Agents on platforms like @youdotc3.4×59k41340145
55Nov 26, '24Just 6 years ago. This NLP reviewer was 100% certain that prompt engineering to unify all NLP problems in a single neural network and just ask it any question was completely misgui3.3×58k61249151
56Jul 19, '25Fun surprising science fact for your Saturday morning. Liquids don't splash when hitting a hard surface in a vacuum. Somehow nobody really knew that so researchers actually wrote a3.2×56k51937115
57Oct 30, '23Keep AI open.3.1×55k354573
58Mar 20, '24It's nice they covered my thoughts. I learned that the journalist who writes the article is usually not in charge of its title. That's done by their editor who only thinks about cl3.1×55k3572333
59May 11, '17A new reinforcement and deep learning model for summarization.It generates long language sequences that make sense.3.1×06773621
60Mar 14, '25This strengthens my hypothesis that there is a large set of equivalent neural architectures of which the original transformer is just one sample.3.1×54k56246173
61Dec 8, '20Let's build a new internet together.3.0×085710363
63Jul 14, '22So excited to announce our 25m Series A to build a better and more open search engine. We will focus on * youCode - the search engine for developers and * opening up the platform f2.9×01.2k4335
64Mar 10, '17Rather than spending a month figuring out an unsupervised machine learning problem, just label some data for a week and train a classifier.2.8×063127611
65Oct 29, '19Summarization is one of the most important & least solved tasks in #NLProc Problem with all #DeepLearning models: they are not optimized for factual correctness We introduce a 2.7×062821593
66Jul 2, '19One of the hardest problems in #AI is common sense reasoning. This paper by @nazneenrajani, @BMarcusMcCann, @CaimingXiong+I makes huge progress on this. Powerful, simple and unsupe2.6×060820099
67Aug 28, '18A new #deeplearning model for improving abstractive summarization by actually creating novel phrases. Reinforcement learning in #naturallanguageprocessing. Work by amazing intern @2.5×058622765
68Nov 7, '23Most cost efficient way to try out #GPT4 is you . com right now. Plus it's connected to the internet and up-to-date, has the most accurate citations for facts, personalizes your an2.4×42k6113517
69Jan 16, '24It's kind of absurd to insinuate that open source AI is more dangerous than any other kind of AI. The more you can study something, the more you can understand it and think of coun2.4×42k3014322
70Mar 11, '22I've worked on academic deep learning and summarization for years. Summarization is a foundational technology for the information age and a remedy for the attention economy. Here's2.4×053169211
72Oct 24, '23I was honored to be invited to lead a session on AI and Europe with @Bundeskanzler Olaf Scholz and the French President @EmmanuelMacron and both of their cabinets. We covered a lot2.2×39k327167
73Jan 11, '22We built so you can search less and do more. To take back control of your information diet and privacy. Here are 7 tips to get the most out of our search engine. 🧵👇2.2×06407488
74Jan 14, '19Three road blocks to solve for artifical general intelligence (#AGI) 1) Massive Multitask learning with a single joint model 2) Ability of algorithms to update their objective func2.0×055313435
75Aug 6, '20Let's build an #AIEconomist for the real world together! The framework for building RL-friendly economic simulations is now open source! This has so much potential for good.2.0×050212080
76Dec 21, '17A reinforcement learning agent that learns to program new neural network architectures. Same/better results as LSTMs but with funky nonlinearities (sine, SeLus, etc) and new connec2.0×04962052
77Aug 9, '17Excited to announce that I'm now an adjunct professor at the #Stanford computer science department. 🙃 Still chief scientist at #Salesforce1.9×0743470
78Nov 9, '21Today, we launched , a private search engine that summarizes the web for you 🚀. I'll be answering questions throughout the day on @ProductHunt. Thank you for your support and feed1.9×04819376
79Aug 17, '22Introducing the Social Tab on : search 9+ social media apps at once. A 🧵 on features that let you find people, places, ideas, travel inspiration & how-tos fast across: 🐙 Redd1.9×07362833
80Dec 20, '17Hierarchical and interpretable multi-task reinforcement learning. This tackles some major limitations in RL: Training agents to solve complex problems that require multiple subtask1.9×05131752
81Sep 3, '19The calm before the storm. This feels like a once in a life time shot. The probability for everything to align was tiny. The result? A gorgeous sleeping leopard in an acacia tree w1.8×06308112
82Oct 6, '22Is there a more useful search engine out there with cooler AI features than ? I think not (: Now with a beautiful AI image generation app:1.8×06813436
83Mar 15, '22Today I'm excited for another step from a search engine to a do engine: , an AI-powered writing assistant. If you are a copy writer, marketer, student, emailer, or somehow stuck on1.7×04485594
84Nov 7, '17First fully parallel machine translation model.No RNNs😳Less latency🙃1.7×04111721
85Jul 13, '20Personal update: After 4+ amazing years, I will leave my chief scientist role at Salesforce and transition to be an advisor. Proud of what we have accomplished with the Einstein te1.6×06402811
86Jul 25, '18website is up! Slides motivating true multitask learning in AI and NLP from a recent talk:1.5×035214024
87Sep 20, '21Searching for code snippets and questions can be frustrating and time-consuming. Until now: will help you solve your coding questions faster and summarize the relevant web for you.1.5×03497383
88Jun 29, '22So excited to launch our developer program to build apps on I think it's so important for the next search engine to be much more open, let everyone collaborate, help users actually1.5×05505010
89Sep 9, '20Do you want to help build the next defining internet consumer company on values like trust, facts, kindness and equality? We're hiring our first team members for product, UI, UX, f1.4×04065657
90Jun 12, '20In a surprising twist/pivot, OpenAI launches a waitlist for a neat enterprise software API. @Huggingface becomes an open ai company with amazing open source nlp toolkit. I'm loving1.4×05165021
91Nov 8, '17My TEDx talk on the present and future of artificial intelligence is online1.4×03691253
92Sep 14, '18My keynote about the future of natural language process, #deeplearning, multitask learning and For the remote conference AI with the best1.4×032210744
93Aug 6, '18Prediction: Any AI problem that you can simulate and sample endlessly many training samples for can be solved with today's algorithms such as deep and reinforcement learning.1.4×03519220
94Mar 11, '20Introducing ProGen, a large language model trained on 280 million protein sequences that can generate viable proteins based on user specifications. A step towards AI & #nlpproc1.3×03319930
95Jul 31, '19Best Paper Award at @ACL2019_Italy! The paper improves dialogue systems through transfer and few-shot learning. Congrats to our co-authors and our intern and future member of the @1.3×03907429
96Jul 31, '19Can you overfit on your data? —> Collect more training data. Do you underfit your data? —> Train a bigger model. It is surprising how many neat research results came from thi1.2×03777922
97Nov 6, '19Hypothesis: any general function approximator that is powerful enough (eg has enough parameters) and can be efficiently trained will eventually perform similarly. The underlying su1.1×03265628
98Aug 1, '20This was possibly my most epic flying moment. So much beauty but also fear. Hope you'll enjoy it as much as I did (wait a few seconds for the big reveal). #iceland #paramotor #wate1.0×0385257
100Jul 19, '22Today we will answer questions on about YouCode, a developer search engine from @YouSearchEngine. Code and learn up to 5x faster 🧑‍💻 ⚡ Search 20+ apps for coding efficiency witho0.9×0359265

Scan another creator

@

Run this on any account.

XOutlier builds this table for every public profile you open on x.com, with live data and every post, not just the top 50.