"It is hard to be wildly successful at anything you aren't obsessed with." -- Sam Altman https://t.co/ogp6cJ8MNj
— Z Fellows (@ZFellows_) Dec 30, 2023
from Twitter https://twitter.com/ZFellows_
December 30, 2023 at 01:24PM
via IFTTT
"It is hard to be wildly successful at anything you aren't obsessed with." -- Sam Altman https://t.co/ogp6cJ8MNj
— Z Fellows (@ZFellows_) Dec 30, 2023
The Human Side of Decision Making by Daniel Kahneman. I highly recommend you read it. A quick 10-pager. Here's a free PDF: https://t.co/ccHq7Eo4Bz https://t.co/8udptFOShQ
— Daniel (@MnkeDaniel) Dec 29, 2023
New Bishop "Deep Learning" book is available free at the bottom of this page. https://t.co/7SDDZYd7So (Although I bought a copy, because I need to convey gravitas)
— Sasha Rush (@srush_nlp) Dec 29, 2023
I expect to see many more companies offer performance-based pricing. Historically most SaaS companies were tools used by humans to accomplish a goal. As AI takes on more of the burden – vertically integrating the process from customer outreach to solution – new pricing models… https://t.co/dYcz0dXXzj https://t.co/DDGEx5D6sb
— Ryan Hoover (@rrhoover) Dec 29, 2023
re-reading @VitalikButerin's excellent post from last night and struck by something I've been always puzzled/saddened by. crypto descends from 1990s cipherpunk/FOSS lineage. It should be directly a sibling of modern OSS / privacy institutions: Mozilla, Wikimedia, Signal, etc.… https://t.co/TTO6NgggOm
— Sriram Krishnan - sriramk.eth (@sriramk) Dec 29, 2023
SqueezeLLM is a post-training quantization framework that incorporates a new method called Dense-and-Sparse Quantization to enable efficient LLM serving. π₯ π Dense-and-Sparse splits weight matrices into two components: A dense component that can be heavily quantized without… https://t.co/IQBfB9vpjR https://t.co/WrGvbhVBi3
— Rohan Paul (@rohanpaul_ai) Dec 29, 2023
Meet our newest Wander locations. In the past week, we launched three new Wanders in Big Sur, California, Kure Beach, North Carolina and Big Sky Montana. We have one more coming this weekend before 2023 ends. Where do you think it is? https://t.co/XSlvBj3kvD
— Wander (@wander) Dec 29, 2023
Statistics 110: Probability by Harvard University. Inarguably, one of the best probability courses on the web! https://t.co/TJfLjT6ISn https://t.co/Dwk6QsLdY8
— Sumanth π (@Sumanth_077) Dec 28, 2023
To create the perfect user interface, develop it in a fast feedback loop where the feedback is provided by an 8-year-old.
— FranΓ§ois Chollet (@fchollet) Dec 29, 2023
I hate wasted potential When I see someone being forced to take less than they deserve Or accepting less than they can get It breaks my ❤️ I wrote Burn the Boats to right those wrongs π₯⛵️ Pls let me know if it helped you become more of you π https://t.co/Q9S9f1RzOn
— Matt Higgins (@mhiggins) Nov 22, 2023
Disney, WBD, Comcast, and Paramount face a year of reckoning in 2024 after losing $5B+ in 2023 from the streaming services they built to compete with Netflix (Financial Times) https://t.co/oIvY8ptgVI π« Subscribe: https://t.co/OyWeKSRpIM https://t.co/L8sxTX03sf
— Techmeme (@Techmeme) Dec 28, 2023
Really enjoying experimenting with Gemini Pro on Google AI Studio. The example below shows how to create a few-shot prompt for a simple emotion classification use case. It allows you to create so-called "structured prompts" where you can add exemplars for the few-shot prompt in… https://t.co/ETZzth2ddh https://t.co/ta7qQJynw4
— elvis (@omarsar0) Dec 28, 2023
π¦LangChain x Mixtral x TogetherAI This community blog post takes a deep dive into comparing Mixtral performance vs GPT-3.5 on 15 questions of reasoning, logic, and coding Great example of how to use it in LangChain via @togethercompute Blog: https://t.co/q81lsWj8WN
— LangChain (@LangChainAI) Dec 28, 2023
Keith Rabois on how to identify great talent “What you want to do with every single employee every single day is expand the scope of their responsibilities until it breaks… and that’s the role they should stay in.” In the clip below, Keith tells the story of giving an intern… https://t.co/IBRXxNmteK https://t.co/Het6LKHAS1
— The Startup Archive (@StartupArchive_) Dec 28, 2023
> Book list for streetfighting computer scientists > do you need to read all of these? only if you want to know the things inside, i guess. https://t.co/NayTgIA4t3 https://t.co/tFu3IeJh6V
— Phil Eaton (@eatonphil) Dec 28, 2023
πΈ️Customized LangChain <> @neo4j Integration If you want to build advanced RAG, you probably need to know how to customize any off-the-shelf guide to your use case @tb_tomaz dives deep on how to do this with Graph DBs like Neo4j Blog: https://t.co/xZGRM1DF8M https://t.co/htH41K8hUj
— LangChain (@LangChainAI) Dec 27, 2023
"Using LLMs to detect the level of hallucinations in the responses of other LLMs." Sounds odd, but there is already promising research: •G-Eval •GPTScore •SelfCheckGPT •TRUE •ChatProtect •Chainpoll I am curious about your thoughts on using LLMs to quantify hallucinations. https://t.co/iPu1GWNIvY
— Leonie (@helloiamleonie) Dec 27, 2023
The cycle of high-performing teams: - Insatiable curiosity - Strong bias for action - Challenge & feedback - Continual improvement Great leaders power the cycle by: - Holding high standards - Developing the talent - Removing friction - Being decisive The fastest team wins.
— Dave Kline (@dklineii) Dec 27, 2023
10 of the Best Company Engineering Blogs I read this year: 1) How Discord stores trillions of messages: https://t.co/KdukyqRQJ9 2) Real time messaging at Slack: https://t.co/0Z5z2P27l4 3) How Canva scaled Media uploads from Zero to 50 Million per Day: https://t.co/q9pMs05mXo… https://t.co/oVd04g86RM
— Ashish Pratap Singh (@ashishps_1) Dec 27, 2023
11) The Gray Lady Winked by Ashley Rindsberg I give this book by @AshleyRindsberg 10/10 recommendation. Everybody in crypto should read it. I put this up there with The Sovereign Individual. https://t.co/5YEKhTThjT
— Reads with Ravi (@readswithravi) Dec 26, 2023
Lots of you have asked me how to train a custom Stable Diffusion model to generate consistent characters. Just found a really easy way to do it. You can even use it to generate images of yourself. Here’s how:
— Chase Lean (@chaseleantj) Dec 26, 2023
Reading @kaifulee AI 2041. The book is a great showcase of how insane Ai progress is: the majority of Kai-Fu's predictions for 2041 are happening now and will become a long gone history by the time 2041 rolls around. https://t.co/uhxYGsxbfc
— Starson (@DrStarson) Dec 26, 2023
π’Thrilled to announce Intel Extension for Transformers v1.3 released, featuring 1) efficient low-bit inference and fine-tuning, and 2) improved open-source chatbot framework Neural Chat. π¨π»Notes: https://t.co/opPoeAQ5Vt π€Code: https://t.co/oWYnKP7mG1 X'mas and Happy New Year!
— Haihao Shen (@HaihaoShen) Dec 25, 2023
2023 was AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI Biggest AI releases this year! https://t.co/7XtEntTTMK
— Shubham Saboo (@Saboo_Shubham_) Dec 25, 2023
@sriramk Peter F Hamilton's Salvation trilogy is an absolute masterpiece of big, epic, far future sci-fi with a war thriller rolled into one.
— Daniel Jeffries (@Dan_Jeffries1) Dec 24, 2023
Lots of similarities IMO - <s>Annoying data janitoring</s> - Metric & eval design - Fine tuning: pipelines, ops, infra - Error analysis - Debugging model perf - Being skeptical of everything https://t.co/XIgRXXmoNL
— Hamel Husain (@HamelHusain) Dec 24, 2023
My kingdom for an analysis of "what went wrong?" that doesn't simply ignore one, or several, massive elephants sitting right there in plain sight.
— Marc Andreessen πΊπΈ (@pmarca) Dec 24, 2023
@pmarca Probably: “There were humans in charge of stuff who got there without earning their power and made decisions about situations where they had no skin in the game.” Coupled with: “There was sufficient prosperity that people got complacent about where it came from.”
— Michael J South (@renegademormon) Dec 24, 2023
LLM Interpretability: The unsung hero of AI advancements. Delve into this essential yet overlooked field with a repository that's a one-stop-shop for groundbreaking tools and papers. https://t.co/xswK5wp0T6
— Rohan Paul (@rohanpaul_ai) Dec 24, 2023
Holidays are the perfect time to relax with a captivating book! At Turing Post, we've handpicked a selection of books personally recommended by AI&ML experts. Here's an exclusive sneak peek at some fresh releases from 2023! π§΅
— Ksenia Se (@Kseniase_) Dec 24, 2023
For those who enjoyed the below piece (which is excellent), this interview with Janssen is also great: https://t.co/JSlgaaVB8h. https://t.co/fZneObuYMi
— Patrick Collison (@patrickc) Dec 24, 2023
Important paper from the inventors of StyleGAN https://t.co/hSGcN9FOKf https://t.co/QDYuJcm0L8
— Aravind Srinivas (@AravSrinivas) Dec 23, 2023
Elon Musk: "Failure is irrelevant unless it is catastrophic." https://t.co/go8uC3PZ2V
— BUILD OR DIE (@BUILD_OR_DIE) Dec 23, 2023
Sources: OpenAI is in early talks to raise new funding at a valuation at or above $100B, and discussed raising between $8B and $10B from G42 for a chip venture (Bloomberg) https://t.co/rrbFTky9Xm π« Subscribe: https://t.co/OyWeKSRpIM https://t.co/O1oGIQBbOg
— Techmeme (@Techmeme) Dec 22, 2023
12 thoughts on incentives: 1. Don't ask your barber if you need a haircut. 2. "I can fix the $32 trillion US debt problem in 5 minutes. You pass a law that when there’s a deficit of more than 3% of GDP, all sitting members congress are ineligible for re-election” - Warren… https://t.co/VnfbpFBf5w https://t.co/qV4uFb9upD
— George Mack (@george__mack) Dec 23, 2023
LLM Interpretability hasn't received as much attention but it's as important as any other research area to keep advancing LLMs. This area of study has implications for how to optimize and make LLMs more performant, reliable, and safer. I really like this repo that curates… https://t.co/9itQ3m5jWx https://t.co/ftN86VaQXh
— elvis (@omarsar0) Dec 23, 2023
Big AI discovered a new structural class of antibiotics! The last one took 38 years with multiple compounds effective vs methicillin-resistant Staph aureus, without toxicity. Article: https://t.co/ZxEVr8g1zb
— Brian Roemmele (@BrianRoemmele) Dec 22, 2023
Eigenvalues & Eigenvectors clearly explained! The concept of eigenvalues & eigenvectors is widely known yet poorly understood! Today, I'll clearly explain their meaning & significance. Let's go! π https://t.co/FoKM3qccyb
— Akshay π (@akshay_pachaar) Dec 22, 2023
1/ @ammaar https://t.co/9KTCvh1pwy
— Min Choi (@minchoi) Dec 22, 2023
Multimodal Agents as Smartphone Users Introduces an LLM-based multimodal agent framework to operate smartphone applications. Learns to navigate new apps through autonomous exploration or observing human demonstrations. Shows proficiency in handling diverse tasks across… https://t.co/mYpZr55sXr https://t.co/CiJmAf0nkf
— elvis (@omarsar0) Dec 22, 2023
“Plans should be measured in decades, execution should be measured in weeks.” π https://t.co/vmFAHz6afx
— Jim Fan (@DrJimFan) Dec 22, 2023
Been quite a year Thanks to all of you for being alongside us on the journey π π for a great 2024! https://t.co/l2p3NNAF4h
— Shomik Ghosh (@shomikghosh21) Dec 22, 2023
Hearing multiple reports that the girls are exporting their WhatsApp messages and taking them to ChatGPT for relationship analysis
— Justine Moore (@venturetwins) Dec 22, 2023
RAG for LLMs Great to see an overview of all the retrieval augmented generation (RAG) research that has been happening. Should be a great read for the end of the year. https://t.co/criuHgH5E8 https://t.co/nYiWuQZKdP https://t.co/E8LJPLGHWy
— elvis (@omarsar0) Dec 23, 2023
Benford’s law of controversy: Passion is inversely proportional to the amount of real information available. Applies to so many things.
— Morgan Housel (@morganhousel) Dec 21, 2023
✨ Just launched: A digital archive of all Buffett's letters to shareholders ✨ π Search the letters by topic, question meaning, or similarity π Browse each letter & search for similar sections (link in thread) https://t.co/Rj5zkFqVZr
— Max Olson (@maxolson) Dec 21, 2023
LLMs will eat software Here's a nice view of the larger market segments. In just 1 year: > OpenAI went from 0 to $1B+ ARR > Nvidia data center revenue went 6x from $3.6B to $18B+ Source: IOT Analytics (dm for the link to the report) https://t.co/Qy1RNr4LwJ
— Chief AI Officer (e/acc) (@chiefaioffice) Dec 21, 2023
what i wish someone had told me: https://t.co/1nEaYimzXG
— Sam Altman (@sama) Dec 21, 2023
The 2023 Stratechery Year in Review https://t.co/RSR6Hxjkdl
— Stratechery (@stratechery) Dec 21, 2023
π¨“Buy Now, Pay Later” (BNPL) now being offered at thousands of Walmart stores through self-checkout This year, BNPL hit an all-time high on Cyber Monday up +42.5% YoY Expected to be a record year for instalment payment providers, according to Adobe This is happening during a… https://t.co/C76A07Lknb https://t.co/XbBn2XtGt2
— Genevieve Roch-Decter, CFA (@GRDecter) Dec 20, 2023
The Gemini report is now up on Arxiv. https://t.co/mz8tFMTBB7 https://t.co/gYqnzIbrX0
— Jeff Dean (@π‘) (@JeffDean) Dec 20, 2023
The highest revenue multiple companies spend 1.6x on S&M than R&D But the more spent on S&M isn’t better…π§΅ https://t.co/kRJiYoSyog
— OnlyCFO (@OnlyCFO) Dec 19, 2023
Steve Jobs: "We don't ship junk." https://t.co/Ik4P392JJ7
— Z Fellows (@ZFellows_) Dec 19, 2023
Ikigai (ηγη²ζ) is a Japanese concept referring to something that gives a person a sense of purpose, a reason for living. Japanese have basically a word for any concept that in other languages require lots of words (except German, obviously). The term compounds two Japanese… https://t.co/YbkYByO1OY https://t.co/nXoFhrtMMN
— Massimo (@Rainmaker1973) Dec 19, 2023
Prototyping a free AI youtube summarizer that can run quickly and accurately on any computer. I have a ton of valuable long form (2hr +) youtube videos on my watch list (example below is Jeff bezos' recent interview with Lex Fridman), but can never find time to watch them. I… https://t.co/ukaomXOqzm https://t.co/8Nn6lYwsrb
— Mayo Oshin (@mayowaoshin) Dec 18, 2023
"We've been working on a tablet project, which had a little odd beginning. It began because Steve hated this guy at Microsoft - that is the actual origin." — Scott Forstall, about the iPad, 2017 https://t.co/I1U4qQ5hJ3
— Steve Jobs Stories (@SJobs_Stories) Dec 18, 2023
Successful people surround themselves by supportive truth tellers https://t.co/YHemH3ECUt
— Vala Afshar (@ValaAfshar) Dec 17, 2023
DataComp (https://t.co/8pOYdld89z / https://t.co/ToDD8xCP3V) Oral at #NeurIPS2023 - showing the power of developing data for LLMs. Tl/dr: Fix all the model training code; develop (sample/filter) the pre-training data. New SOTA! Models are commodity- curated data is the edge. https://t.co/4anMLs2TW7 https://t.co/1qXjwD3zn3
— Alex Ratner (@ajratner) Dec 17, 2023
Retrieval-Augmented Generation (RAG) is revolutionizing the field. Introduced by Meta AI researchers in 2020, it's gaining attention for its remarkable advantages. Let's dive into the world of RAG: https://t.co/MkIU9MtDAD
— Ksenia Se (@Kseniase_) Dec 17, 2023
New @DataScienceHarp piece dissects how WebBaseLoader, SeleniumURLLoader & NewsURLLoader in @LangChainAI π¦️π shape #RAG model performance. Some surprising context recall vs precision tradeoffs! Click to explore key metrics & optimization tips. https://t.co/5ey6ca8Rwo #LLMs https://t.co/LUJR2uPghJ
— Roger Taylor (@Roger_M_Taylor) Dec 17, 2023
In this long interview, Jeff Bezos indicates the shift in his attention from the internet to space. I suspect we'll be hearing more about Blue Origins in the coming years. https://t.co/LANk5bdBJj
— Kevin Kelly (@kevin2kelly) Dec 17, 2023
Steve Jobs on avoiding “the bozo explosion” (companies fill up with second-rate talent as they grow) "For most things in life, the range between best and average is 30% or so. The best airplane flight, the best meal, they may be 30% better than your average one. What I saw… https://t.co/l33nqoXAxl https://t.co/KegbRPOvO9
— David Senra (@FoundersPodcast) Dec 17, 2023
For years I told friends that are researchers at OpenAI, Deep Mind, others that AI Hallucinations are not only vital in some cases but one day, but encouraged. We are just now seeing papers that are showing critical new discoveries that show new emergent creativity from this. https://t.co/cLmDwphzYm
— Brian Roemmele (@BrianRoemmele) Dec 17, 2023
In 2024, you will see the ad industry reap profits from Generative AI. Hyper-personalized tailor-fit ads per user, GenAI ad creative, and increased ROAS because of better AI targeting will result in higher CPMs and, later on, increases in ad budgets. Watch out for $META &… https://t.co/AkQnPpJXF0
— Rihard Jarc (@RihardJarc) Dec 16, 2023
The Animate Anyone saga continues! DreaMoving is yet another approach at generating high-quality videos of humans given a text prompt and some pose guidance. In this case, a reference image is used to preserve facial identity. https://t.co/Lak860Ps2N https://t.co/grPboNLUOH
— Dreaming Tulpa π₯π (@dreamingtulpa) Dec 16, 2023
This is crazy! π€―π€―π€― It's free, super easy to use, and accessible to everyone... More info in the first comment! https://t.co/oQvgxiEgud
— TechHalla (@techhalla) Dec 15, 2023
New direction for AI alignment — weak-to-strong generalization. Promising initial results: we used outputs from a weak model (fine-tuned GPT-2) to communicate a task to a stronger model (GPT-4), resulting in intermediate (GPT-3-level) performance. https://t.co/mX2NOOwhQ4
— Greg Brockman (@gdb) Dec 14, 2023
Open AI released a great Prompt Engineering guide (link in comments). Here is a summary of their 6 strategies for getting better results when prompting GPT-4 https://t.co/QiBASsbOHD
— MindBranches (@MindBranches) Dec 15, 2023
Top 7 @JeffBezos quotes from the latest @lexfridman episode: 1. “We invent new tools and then our tools change us.” 2. “Long-term thinking is a giant lever. You can literally solve problems if you think long-term that are impossible to solve if you think short-term.” 3. “Be… https://t.co/8fXjFWhMe8 https://t.co/oLpiCJxMXF
— Podcast Notes π️ (@podcastnotes) Dec 15, 2023
What do marketplaces look like in the age of AI? Some will grow, some will shrink, and some will disappear. It's a new era for the marketplace model! My take on who will be impacted - and how π https://t.co/uibIJR0HnN
— Olivia Moore (@omooretweets) Dec 14, 2023
FYI to anyone using @MistralAI's Mixtral for long context tasks -- you can get even better performance by disabling sliding window attention (setting it to your max context length) config.sliding_window = 32768 https://t.co/K53Mwfc519
— emozilla (@theemozilla) Dec 14, 2023
We can automate everything with AI Agents, we can analyze everything with LLMs. Human expert predictions/reviews with LLMs + predictive analysis with ML = πΈππ½ In @LangChainAI, @togethercompute, @MistralAI We Trust Accelerate by delegating... https://t.co/mtqNBtPjrm
— Muratcan Koylan (@youraimarketer) Dec 14, 2023
Today developers can start building with our first version of Gemini Pro through Google AI Studio at https://t.co/ozfVwuBpSZ. Developers have a free quota and access to a full range of features including function calling, embeddings, semantic retrieval, custom knowledge… https://t.co/VP7sYufzy2 https://t.co/GHzFaoU9cK
— Sundar Pichai (@sundarpichai) Dec 13, 2023
Completely AI-generated news anchors are going VIRAL. Plus, huge developments in AI from Tesla Optimus, Microsoft, ChatGPT, Anthropic, and 8 new AI tools. Here's the rundown of everything you need to know:
— Rowan Cheung (@rowancheung) Dec 13, 2023
'Channel 1' just revealed their new AI news anchors. All avatars and voice in this video is AI-generated. With AI video tech improving week over week- the line between what's real and AI-generated is only going to continue to blur. https://t.co/eLTaoIe1i7
— Rowan Cheung (@rowancheung) Dec 13, 2023
Marc Andreessen: "The world is a very malleable place. If you know what you want, and you go for it with maximum energy and drive and passion, the world will often reconfigure itself around you much more quickly and easily than you would think." https://t.co/8gll4JeiWU
— BUILD OR DIE (@BUILD_OR_DIE) Dec 12, 2023
There's too much happening right now, so here's just a bunch of links GPT-4 + Medprompt -> SOTA MMLU https://t.co/TQ8KgUBsnQ Mixtral 8x7B @ MLX nice and clean https://t.co/YpWNuQZgxA Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models… https://t.co/2IJbdDQifF
— Andrej Karpathy (@karpathy) Dec 12, 2023
Good morning, friends π ! Start your day with a morning #prompt in biomechanical translucent style to enhance objects or characters. an image of a biomechanical translucent [Your object], with golden lights, in the style of vray tracing, polished craftsmanship, fluid organic… https://t.co/ydgDIZU3JB https://t.co/g6Kyt2lLY0
— Pierrick Chevallier | IA (@CharaspowerAI) Dec 12, 2023
Man, this open-source AI chat is almost too good to be true. And it's MIT licensed π€―. https://t.co/HahwiY5GrT
— Minh-Phuc Tran (@phuctm97) Dec 12, 2023
RAG App Slow & Costly? Smarter LLM Chunking With @LangChainAI by ClΓ©ment ROBIN fixes that! He uses LangChain for an innovative new strategy that intelligently summarizes docs with LLM first, then chunks into coherent bits for better context. Read more! π https://t.co/VmKrFpsDjp https://t.co/zju9kF2o4d
— Roger Taylor (@Roger_M_Taylor) Dec 12, 2023
Official post on Mixtral 8x7B: https://t.co/ce0ZjHhLVn Official PR into vLLM shows the inference code: https://t.co/vJbmDG9RhG New HuggingFace explainer on MoE very nice: https://t.co/lTaNCONUeI In naive decoding, performance of a bit above 70B (Llama 2), at inference speed… https://t.co/WE9rp74c3D https://t.co/OMSTfYXVsE
— Andrej Karpathy (@karpathy) Dec 11, 2023
“Use LLMs as a reasoning engine to process information, rather than using it as a source of memorized information” Great tip by Andrew Ng. https://t.co/SAaOqAwUNv
— elvis (@omarsar0) Dec 11, 2023
You can use Dall-E 3 in any app on your phone. It works on iOS and Android. 1. Download and activate Microsoft Swiftkey keyboard (link below) 2. Tap on the emoji then on the Designer icon 3. Write your prompt and insert your image Note: it generates 4 different images… https://t.co/egX42uoJU0 https://t.co/26u9Lbbk7x
— Paul Couvert (@itsPaulAi) Dec 11, 2023
ππΆπΏπ―π»π―'π π πΆπ°πΏπΌππ²πΏππΆπ°π² ππΏπ°π΅πΆππ²π°πππΏπ² Jessica Tai, an engineering manager at Airbnb, recently talked at QCon about how Airbnb's architecture changed over the years and followed company development. It went through the following phases: π.… https://t.co/tXTKZ8UYLF https://t.co/0Ox92ujgYD
— Dr Milan MilanoviΔ (@milan_milanovic) Dec 11, 2023
Proud to announce: Mixtral 8x7B -- Mixtral of Experts - Free to use under Apache 2.0 license - outperforms Llama 2 70B with 6x faster inference. - matches or outperforms GPT3.5 - masters English, French, Italian, German and Spanish. - seq_len = 32K https://t.co/mRMQYtxmx6 1/N https://t.co/XMK8sA1KLH https://t.co/BnLzE7fXzR
— Devendra Chaplot @ #NeurIPS2023 (@dchaplot) Dec 11, 2023
Ted Chiang has one of the most profound articles, I've read, on explaining LLMs. He mentions that understanding and compression are two sides of the same coin.πͺ And interestingly, when we’re dealing with predicting words, lossy compression looks smarter than lossless… https://t.co/vpWVPdZcb3 https://t.co/Bm44f2xk3I
— Zain Hasan (@ZainHasan6) Dec 10, 2023
If you're an AI engineer: we're releasing a brand-new cookbook π§π³ highlighting three ways you can build LLM apps by integrating multi-modal models (LLaVa) into the rest of your workflow: 1️⃣ Retrieval Augmented Image Captioning: Caption an image with LLaVa, then augment response… https://t.co/X5iMuvcTgg https://t.co/xXs5hsd9Rl https://t.co/dHvLZT3dvZ
— Jerry Liu (@jerryjliu0) Dec 10, 2023
π¨FRACTAL IMPRESSIONISMπ¨ Base prompt : [SUBJECT] in Fractal Impressionist style, featuring [COLOR] and [COLOR] recursive details As usal play with settings some exemples in ALT https://t.co/vowAQUQ0mM
— LudovicCreator (@LudovicCreator) Dec 10, 2023
Current project: a platform where all the videos are generated using open-source AI models π¨π¬ From a single text prompt it generates the story, snippets, waveforms, merges everything into a single mp4 then pushes it to a Hugging Face dataset Why? Well, because it's fun!π https://t.co/v0mHiAjVRd
— Julian Bilcke (@flngr) Dec 10, 2023
There are two types of mistakes: mistakes of ambition and mistakes of sloth. The first is the result of a decision to act—to do something. This type of mistake is made with incomplete information, as it’s impossible to have all the facts beforehand. This is to be encouraged.… https://t.co/7pn1YntQ02
— Tim Ferriss (@tferriss) Dec 10, 2023
The top ML Papers of the Week (Dec 4 - Dec 10): - Gemini - EfficientSAM - Chain of Code - LLMs on Graphs - Data Management For LLMs - The Efficiency Spectrum of LLMs ...
— DAIR.AI (@dair_ai) Dec 10, 2023
A critique of (the way people understand) Tree of Thoughts (ToT) paper: #SundayHarangue From the time the ToT paper came out on arXiv, the LLM fan club saw it as a savior for lifting LLMs from their System 1 life to System 2 reasoning status. (c.f. https://t.co/YNtXiVLxcy ).… https://t.co/UKS1usVZT7 https://t.co/MckP5e9BBp https://t.co/fRju7WYyWp
— Subbarao Kambhampati (ΰ°ంΰ°ంΰ°ͺాΰ°ి ΰ°Έుΰ°¬్ΰ°¬ాΰ°°ాΰ°΅ు) (@rao2z) Dec 10, 2023
Large language models (LLMs) and knowledge graphs (KGs) are complementary technologies that balance each other's strengths and weaknesses when combined: - LLMs have a strong capability for understanding and generating natural language, but can sometimes hallucinate facts. -… https://t.co/9tnqXiZHqK https://t.co/0b04X0jqoT
— Carlos E. Perez (@IntuitMachine) Nov 5, 2023
"If we realized the pain and suffering and how vulnerable you’re going to feel, the challenges that you’re going to endure, the embarrassment and the shame and the list of all the things that go wrong, nobody in their right mind would do it.” https://t.co/N3MFesA1Yc
— Scott Kupor (@skupor) Dec 9, 2023
Everyone is doing RAG on unstructured docs, but what if you want to mix in structured business data? Databricks RAG can connect to feature tables & functions to query the latest data in your catalog, all with centralized governance, security and MLOps. https://t.co/q8tcfwPJmF
— Matei Zaharia (@matei_zaharia) Dec 9, 2023
important leadership skill is precisely communicating your intent — that is, proposing what the properties of successful execution would be, while leaving the choices of how to achieve those goals for the team to propose in return
— Greg Brockman (@gdb) Dec 9, 2023
.@rabois: "the team you build is the company you build." https://t.co/TOblKmPvpY
— Z Fellows (@ZFellows_) Dec 8, 2023
EfficientSAM was just released and it's fast! π¨ With 20x fewer params, it is now 20x faster than the original SAM segmentation model, while staying in the same accuracy range. See below for the project page and an interactive @huggingface space to try it out! ⬇️⬇️ https://t.co/wLosM0DDx9
— Alex Carlier (@alexcarliera) Dec 9, 2023
Meta just launched their new AI image generator. But how good is it? I compared it to Midjourney, DALL-E 3 and Adobe Firefly across 10 image categories. Here are the results:
— Chase Lean (@chaseleantj) Dec 8, 2023
I know everyone is excited about Mixtral and the new Hyena models - but @ContextualAI just dropped a pile of cool new models and a new alignment framework https://t.co/Mmn3Vvff3T https://t.co/zRErEbQU4q
— Andrew Carr (e/π€Έ) (@andrew_n_carr) Dec 8, 2023
Sam Altman: "I have yet to meet a slow-moving person who is very successful." https://t.co/6gloDbU9cj
— Z Fellows (@ZFellows_) Dec 8, 2023
Meta has released a new free AI image generator. You can access it with a Facebook or Instagram account. I've been using it for several hours and the detail of the images is really impressive. Here are some sample images and how to access it π§΅ https://t.co/RaTXrqydKV
— Paul Couvert (@itsPaulAi) Dec 7, 2023
Sam Altman on how to accomplish big things: 1. Focus 2. Personal Connections 3. Self-Belief / Conviction https://t.co/QxxdMUWKhj
— BUILD OR DIE (@BUILD_OR_DIE) Dec 8, 2023
This talk by Angela Fan on Llama2 is so good. 30 min, she just tells you all the things. https://t.co/K3DvPJmRUa
— Sasha Rush (@srush_nlp) Dec 8, 2023
Introducing Stormer, a scalable transformer model for skillful and reliable medium-range weather forecasting. Stormer achieves competitive accuracy for short-range, 1–7 day forecasts, while outperforming Pangu-Weather and Graphcast by a large margin for longer lead times. Paper:… https://t.co/70lW00eoIS https://t.co/pqfJAHqhkI
— Tung Nguyen (@tungnd_13) Dec 8, 2023
I’m very excited to share our work on Gemini today! Gemini is a family of multimodal models that demonstrate really strong capabilities across the image, audio, video, and text domains. Our most-capable model, Gemini Ultra, advances the state of the art in 30 of 32 benchmarks,… https://t.co/BTJ9so1Khy https://t.co/sQfxBy9tpT
— Jeff Dean (@π‘) (@JeffDean) Dec 6, 2023
Google just launched Gemini: The ChatGPT Killer π€― Gemini comes in three different variants: - Nano (for Android devices) - Pro (Available in Google Bard) - Ultra (Most powerful LLM, coming next year) Multimodal by Design Google didn't followed the footsteps of OpenAI by… https://t.co/1YY9Er10yd https://t.co/rWYH5H6xnU
— Shubham Saboo (@Saboo_Shubham_) Dec 6, 2023
Google Releases Gemini https://t.co/Nm2kK3vIpG https://t.co/49YKw8fHR7
— /MachineLearning (@slashML) Dec 6, 2023
Summary of Gemini's 60-page technical report. 1. Written in Jax and trained using TPUs. The architecture, while not explained in details, seems similar to Flamigo's. 2. Gemini Pro's performance is similar to GPT-3.5 and Gemini Ultra is reported to be better than GPT-4. Nano-1… https://t.co/SPyqhU4NK0 https://t.co/YhXEeaOo01
— Chip Huyen (@chipro) Dec 6, 2023
Checkmate. ♟️ https://t.co/ZTZBoaUTQc
— Marc Benioff (@Benioff) Dec 7, 2023
Gemini is here! Google DeepMind just announced Gemini, their largest and most capable AI model. A short summary of all you need to know: 1) What it is - Built with multimodal support from the ground up. Remarkable multimodal reasoning capabilities across text, images, video,… https://t.co/yydJYffORt https://t.co/HO8mCYYfvS
— elvis (@omarsar0) Dec 6, 2023
With our brand-new `RAGEvaluatorPack` ✨ you can benchmark your RAG pipeline over any eval dataset in one line of code ⚡️ Huge shoutout to @_nerdai_ for creating this. This integrates seamlessly with our Llama Datasets or your own eval datasets. Metrics π: Correctness,… https://t.co/jBpDKtmzFa https://t.co/HHCk2lqgfr
— LlamaIndex π¦ (@llama_index) Dec 6, 2023
Meta launches Imagine with Meta, a standalone text-to-image generator powered by its Emu model that generates four images per prompt, free for users in the US (@kyle_l_wiggers / TechCrunch) https://t.co/q4YXS7MidW π« Subscribe: https://t.co/OyWeKSRpIM https://t.co/YCeYqyqxZA
— Techmeme (@Techmeme) Dec 6, 2023
A high school student reverse engineered the iMessage protocol and published a cross-platform proof of concept. https://t.co/MjGqXYEVwK
— Yuri Sagalov (@yuris) Dec 5, 2023
LLMs as the New Operating System Imagine an operating system where the core process is an AI brain - that's what Large Language Models (LLMs) are becoming. Karpathy in his new video relates LLMs to the kernel of an entirely new kind of operating system, with a 'context window'… https://t.co/hepFTM177p https://t.co/92SehQUU0d
— Shubham Saboo (@Saboo_Shubham_) Dec 6, 2023
Releasing two updated Tess models (Tess-7B & Tess-34B)-v1.4 today. V1.4 was trained to have improved coding capabilities, Tree-of-Thought, and another 10 new system prompts that I've designed to elicit working memory, reasoning and logic, philosophical thinking and… https://t.co/y3edB82ydZ https://t.co/u5XuoZPY4S
— Migel Tissera (@migtissera) Dec 5, 2023
For the first time, you can now read Poor Charlie's Almanack for free online. (Nice hardcovers are also available: https://ift.tt/ExN9u7Y) https://t.co/LV60aqoTEP https://t.co/fNuV1Dm5TN
— Patrick Collison (@patrickc) Dec 5, 2023
Generative AI video just got crazy π€― Meet AI modify region by Pika Labs that lets you edit LIVE videos with just a text prompt. Change dress patterns or convert lamp to xmas tree in a matter of seconds using AI. https://t.co/d2NDNNFzGi
— Shubham Saboo (@Saboo_Shubham_) Dec 5, 2023
Incredible way to visualize what's actually happening within an LLM: https://t.co/Bp9e8r8AzO https://t.co/8ZRjfySvP9
— MatthewBerman (@MatthewBerman) Dec 4, 2023
⭐️Building RAG Pipelines that Use Multiple LLMs: Query Transformation ⭐️ Many advanced RAG pipelines make multiple calls to LLMs at various stages. Most existing templates currently use the same LLM for all stages, but there's no reason that has to be the case. Over the coming… https://t.co/RcBKbRl4zt https://t.co/ZOyeK5QU4G
— LangChain (@LangChainAI) Dec 4, 2023
Today we’re introducing Llama Datasets π¦π - a set of community-contributed datasets that allow users to easily benchmark their RAG pipelines for different use cases. It’s hard to define a single right eval dataset for your LLM use case. Instead, you can pick and choose the… https://t.co/iMP7yeVp9I https://t.co/6fw3bZC6s6
— LlamaIndex π¦ (@llama_index) Dec 4, 2023
The Efficiency Spectrum of LLMs 2024 will be the year of LLM efficiency! If you are interested in catching up on methods for improving efficiency in LLMs, including algorithmic and hardware solutions, this is a great survey. There is a lot of work, largely going unnoticed,… https://t.co/UHndjS20Bm https://t.co/BzfV4yBOup
— elvis (@omarsar0) Dec 4, 2023
Dive into the brilliant mind of Silicon Valley's titan @vkhosla, as unraveled by the masterful storytelling of @DavidKirkpatric. Known for his formidable presence and wisdom, this profile is a must-read to get a glimpse of true leadership. ❤️ https://t.co/QeKpjduMud
— Marc Benioff (@Benioff) Dec 3, 2023
Release update Link to the update : https://t.co/bGOJ8jduRG https://t.co/z5dfU9d6tU
— Stelfie the Time Traveller (@StelfieTT) Dec 3, 2023
This is the first pass on the new chapter for ML Engineering: The AI Battlefield Engineering - What You Need To Know https://t.co/lzBvrjBURs This a WIP and your feedback for improvement is always welcome. https://t.co/kk6qqRswyO https://t.co/qairPHpQov
— Stas Bekman (@StasBekman) Dec 3, 2023
"Short-term optimal is rarely long-term optimal. This is something hindsight teaches, but foresight often misses." A Tiny Thought in the @farnamstreet newsletter https://t.co/8o4zh7HNiD
— Mahima Seth (@mahimaseth28) Dec 3, 2023
ML Community loved @karpathy 's introduction to Large Language Models. In the following blogpost, I think out loud about the implementation of Intelligent Operating Systems: The LLM OS. https://t.co/OsJ1TwvhY0 #MachineLearning #LLM #GPT #OperatingSystem #jarvis
— Anshuman Mishra (@anshuizme) Dec 3, 2023
Accountability is not a one-and-done activity. It's an unceasing decision to commit, kickstarting a cycle of sustained growth.
— A.K. | Certified Habit Coach (@AhnnaKeys) Dec 2, 2023
π¦Retrieval-Augmented Generation: From Theory to LangChain Implementation πWe've really been enjoying @helloiamleonie's articles lately Check out this fantastic blog covering the basics of RAG, the theory behind it, and how to use it in practice Blog: https://t.co/OSe5fNBSze https://t.co/lvWHhY8Bx6
— LangChain (@LangChainAI) Dec 3, 2023
π§ͺ Lo-Fi Anime Aesthetic in @runwayml Creating beautiful anime shots is actually super easy in Text-2-Video. π€© All 200 clips from my LoFi mix were created using the following prompt: [Japanese Ghibli Aesthetic] [Prompt] [Vibrant, Optimistic, Lo-fi, Minimalistic] / No Preset https://t.co/rrlWgnUqEJ
— Nicolas Neubert (@iamneubert) Dec 3, 2023
Microsoft's new Florence 2 is big for Computer Vision. It's a merge between Text and Vision. With a single prompt you can instruct the model to do CV tasks like captioning, object detection, grounding, and segmentation. The best part, it only uses a single backbone to handle… https://t.co/Qusyn3zrOF https://t.co/q9LHLkz5Bp
— Lior⚡ (@AlphaSignalAI) Dec 2, 2023
π¦πMistral 7b User Showcase + LangServe & LangSmith This great Austin LangChain 102 meetup produced a fantastic educational video covering OSS models, LangServe, and LangSmith https://t.co/5mfE9r8Zci π @colinmcnamara for hosting, and @Roger_M_Taylor for bringing to our attn https://t.co/64p7O0g684
— LangChain (@LangChainAI) Dec 2, 2023
OpenGPT meets Generative Agents to get Long term Memory π€― Opensource community is once again ahead of the curve, breaking new ground in building custom GPTs. This time, it's about adding long-term memory to OpenGPT. Here's everything you need to know: 1. Specialized Long-Term… https://t.co/z8J0WQCT9Y https://t.co/1B87JcWhBj
— Shubham Saboo (@Saboo_Shubham_) Dec 2, 2023
The best no paywall unfiltered Charlie Munger reading material from the past week: Munger's Three Basic Rules Essay: https://t.co/KUO1a4HmWQ Becky Quick interview transcript: https://t.co/XURURM9haX Stripe will release a 3-hour chat between John Collison and Charlie Tuesday.
— Tren Griffin (@trengriffin) Dec 2, 2023
Sam Altman: "Almost everyone will try to pull you back and say: 'that seems a little bit too crazy, too out there, too ambitious...'" You want to surround yourself with people who make you more ambitious. https://t.co/5zOGu2KZya
— Brian Ji (@BrianJJi) Dec 1, 2023
Creating an AI-Powered Financial Analyst with OpenBB. This will likely be one of my best presentations - the results we are getting are incredible. See it live at Open Core Summit next week! S/o to @MichaelNStruwig π«‘ https://t.co/X7nJFDwNvF
— Didier Lopes (@didier_lopes) Dec 1, 2023
This is the cockpit of one of the most legendary planes ever built, the SR-71 Blackbird. The Blackbird, created by Lockheed Martin's Skunk Works division, was a revolutionary aircraft. Described by a former pilot as "107 feet of fire-breathing titanium," the Blackbird's… https://t.co/99DomOOX1o https://t.co/phWZMANnsp
— Massimo (@Rainmaker1973) Dec 1, 2023
The Evolving Landscape of API Protocols in 2023 This is a brief summary of the blog post I wrote for Postman. In this blog post, I cover the six most popular API protocols: REST, Webhooks, GraphQL, SOAP, WebSocket, and gRPC. The discussion includes the benefits and challenges… https://t.co/zZWLEgdfDE https://t.co/IGulEDgTit
— Alex Xu (@alexxubyte) Dec 1, 2023
Speed, Control, AND Quality Making hyperrealistic AI portraits in 4 mins. with the Krea real-time tool and enhancer https://t.co/ARLnGaSaz6
— titus (@TitusTeatus) Dec 1, 2023
Shocking Crossover: Harry Potter x Star Wars by Pixar, Holiday Special (Not Yet on Disney+) ✍️ Scenario: Human + GPT πΌ️ Image: SDXL / ComfyUI π₯ Video: @pika_labs π️ Voice: @coqui_ai Consistent Characters: IPAdapter & Prompt Interested in discovering the detailed process… https://t.co/eGPvgCdT7P https://t.co/YSkiMF2fAG
— Thibaud Zamora ∞ (@thibaudz) Dec 1, 2023
An extremely well-done segment by @KatieTarasov continues her Chip Giants series on @CNBC. A great inside look at the Apple Silicon journey, and I was delighted to be able to contribute some perspectives. https://t.co/t39ARBmUea
— Ben Bajarin (@BenBajarin) Dec 1, 2023
π¨ Canva just launched Magic Studio. If you haven't tried it, you're falling behind. Here are 12 features that will blow your mind: [ Bookmark to use later π§΅ ] https://t.co/smFuB6nUMW
— Madni Aghadi (@hey_madni) Dec 1, 2023
Streaming now on Runway TV πΊ https://t.co/pyBaAqafvY
— Nicolas Neubert (@iamneubert) Dec 1, 2023
Had an exhausting week? Looking forward to starting a chilled weekend? I've got you! π Check out this 9-minute LoFi mix I made for @runwayml TV! Over 200 clips, all generated within GEN-2. Tune in to Endless Reflections and watch the full video on Runway TV! Link ππΌ https://t.co/RKAJqGxoOu
— Nicolas Neubert (@iamneubert) Dec 1, 2023