41554 blogs · [ { "id": "01a087df-e7b1-707d-813a-410226af7bcc", "title": "Neural Nets Shrink the Planet: Introducing LGSMs", "url": "https://shital.com/blog/tweets/thread/202306071551-neural-nets-shrink-planet-lgsm/", "published_at": "2023-06-07T15:51:00+00:00" }, { "id": "01a087df-e7b1-707d-813a-410226c65f33", "title": "The 26 Million Reasons Apple Thinks Vision Pro Will Soar", "url": "https://shital.com/blog/tweets/thread/202306071309-apple-vision-pro-26-million/", "published_at": "2023-06-07T13:09:24+00:00" }, { "id": "01a087df-e7b1-707d-813a-410226fbc055", "title": "Leaderboard Lies: Broken Metrics from For-Hire Raters", "url": "https://shital.com/blog/tweets/post/202305260917-leaderboard-lies-broken-metrics/", "published_at": "2023-05-26T09:17:31+00:00" }, { "id": "01a087df-e7b1-707d-813a-4102279feeca", "title": "Model Fails Hinton Test: GPT-4 Still Undefeated", "url": "https://shital.com/blog/tweets/post/202305251636-model-fails-hinton-test/", "published_at": "2023-05-25T16:36:57+00:00" }, { "id": "01a087df-e7b1-707d-813a-4102285264f2", "title": "AI Bots Outsmart Google Rankings", "url": "https://shital.com/blog/tweets/post/202305251039-ai-bots-outsmart-google/", "published_at": "2023-05-25T10:39:36+00:00" }, { "id": "01a087df-e7b1-707d-813a-410228f3af61", "title": "Knuth's Verdict on GPT-4: Only 0.01% to AGI", "url": "https://shital.com/blog/tweets/post/202305211535-knuth-gpt4-agi/", "published_at": "2023-05-21T15:35:28+00:00" }, { "id": "01a087df-e7b1-707d-813a-410229376f58", "title": "PaLM2-L's Token Diet: Outsmarting Chinchilla and LLaMA", "url": "https://shital.com/blog/tweets/post/202305180557-palm2-token-diet/", "published_at": "2023-05-18T05:57:09+00:00" }, { "id": "01a087df-e7b1-707d-813a-410229f468ae", "title": "When Gödel Crashes the AI Alignment Party", "url": "https://shital.com/blog/tweets/thread/202305140659-godel-crashes-ai-alignment/", "published_at": "2023-05-14T06:59:36+00:00" }, { "id": "01a087df-e7b1-707d-813a-41022a7a2de0", "title": "When Yellow Fades: GPT-4 Passes Hinton's Paint Test While Bard Falls Short", "url": "https://shital.com/blog/tweets/thread/202305121202-when-yellow-fades-gpt4-passes-hinton-test/", "published_at": "2023-05-12T12:02:24+00:00" }, { "id": "01a087df-e7b1-707d-813a-41022b59306e", "title": "GPT-4 Passes Hinton's Paint Puzzle; Bard Falls Short", "url": "https://shital.com/blog/tweets/post/202305121201-gpt4-passes-hinton-paint-puzzle/", "published_at": "2023-05-12T12:01:45+00:00" }, { "id": "01a087df-e7b1-707d-813a-41022c43bc08", "title": "6B Params and a Dream: GPT-JT Challenges GPT-3", "url": "https://shital.com/blog/tweets/post/202305041424-gpt-jt-challenges-gpt-3/", "published_at": "2023-05-04T14:24:30+00:00" }, { "id": "01a087df-e7b1-707d-813a-41022c9ac278", "title": "Code First, Ask Questions Later: The Understudied Secret Behind GPT-3.5's Success", "url": "https://shital.com/blog/tweets/post/202305010557-code-first-llm-secret/", "published_at": "2023-05-01T05:57:21+00:00" }, { "id": "01a087df-e7b1-707d-813a-41022ca31f54", "title": "LLaMA Distilled: How GPT-4 Fuels Fine-Tuning Ingenuity", "url": "https://shital.com/blog/tweets/post/202304300359-llama-distilled-gpt4/", "published_at": "2023-04-30T03:59:19+00:00" }, { "id": "01a087df-e7b1-707d-813a-41022ca71d34", "title": "PyTorch Smuggles CUDA Onboard", "url": "https://shital.com/blog/tweets/post/202304251456-pytorch-smuggles-cuda/", "published_at": "2023-04-25T14:56:37+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb211d542", "title": "Big Flops, Bigger Bills: GPT-4's $43M Training Cost", "url": "https://shital.com/blog/tweets/post/202304231110-gpt4-flops-43m-training-cost/", "published_at": "2023-04-23T11:10:08+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb3099c7f", "title": "Chill Out Your LLM: Quick Temperature Tuning for Calibration", "url": "https://shital.com/blog/tweets/post/202304120600-chill-out-llm-temperature-tuning/", "published_at": "2023-04-12T06:00:03+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb34d0cd2", "title": "FlashAttention's Oscar-Worthy Performance", "url": "https://shital.com/blog/tweets/post/202304081042-flashattention-oscar-performance/", "published_at": "2023-04-08T10:42:35+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb4226e5c", "title": "FlashAttention: AI's Oscar-Worthy Money Saver", "url": "https://shital.com/blog/tweets/post/202304081037-flashattention-ais-oscarworthy-saver/", "published_at": "2023-04-08T10:37:15+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb478ab65", "title": "Segment Anything vs. The Torture Test: Zero-Shot Showdown", "url": "https://shital.com/blog/tweets/thread/202304060250-segment-anything-zero-shot-showdown/", "published_at": "2023-04-06T02:50:52+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb551a6c0", "title": "AMD MI250x: The Overlooked Powerhouse Outsmarting A100", "url": "https://shital.com/blog/tweets/post/202303290701-amd-mi250x-overlooked-powerhouse/", "published_at": "2023-03-29T07:01:44+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb5b8e841", "title": "Midjourney v5: Hands Up, Letters Down", "url": "https://shital.com/blog/tweets/post/202303271447-midjourney-v5-hands-up-letters-down/", "published_at": "2023-03-27T14:47:26+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb60612cc", "title": "Beam Me Up, Decoder!", "url": "https://shital.com/blog/tweets/post/202303271144-beam-me-up-decoder/", "published_at": "2023-03-27T11:44:19+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb6ecc764", "title": "Be the Prompt, Not the Typist: GPT-4’s Token Triumph", "url": "https://shital.com/blog/tweets/post/202303260008-be-the-prompt-gpt4-token-triumph/", "published_at": "2023-03-26T00:08:45+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb7e38039", "title": "Transformers: More Than Meets the BLEU Score", "url": "https://shital.com/blog/tweets/post/202303251131-transformers-more-than-meets-the-bleu-score/", "published_at": "2023-03-25T11:31:50+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb895425b", "title": "Casually Dropping GPT-4's Trillion Parameters Bombshell", "url": "https://shital.com/blog/tweets/post/202303251054-gpt4-trillion-parameters-bombshell/", "published_at": "2023-03-25T10:54:48+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fb9855981", "title": "Bob Metcalfe Wins 2023 Turing Award, Remains Brutally Honest", "url": "https://shital.com/blog/tweets/post/202303231303-bob-metcalfe-wins-2023-turing-award/", "published_at": "2023-03-23T13:03:49+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fba471c01", "title": "Bard's Brain Drain: Tiny Model, Can't Code, Needs Search", "url": "https://shital.com/blog/tweets/post/202303221336-bards-brain-drain/", "published_at": "2023-03-22T13:36:03+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fba828dc1", "title": "Bard's Coding Conundrum: Smaller Model, Bigger Prompts", "url": "https://shital.com/blog/tweets/post/202303221329-bards-coding-conundrum/", "published_at": "2023-03-22T13:29:16+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fbabf75cb", "title": "Humans: God's Alignment Problem?", "url": "https://shital.com/blog/tweets/post/202303211256-humans-gods-alignment-problem/", "published_at": "2023-03-21T12:56:46+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fbb99772c", "title": "Say Goodbye to the 'L's in LLMs", "url": "https://shital.com/blog/tweets/thread/202303191756-say-goodbye-to-ls-in-llms/", "published_at": "2023-03-19T17:56:24+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fbbe10f62", "title": "The OPT Disasterpiece: LLM Training Gone Wrong", "url": "https://shital.com/blog/tweets/post/202303190655-opt-disasterpiece/", "published_at": "2023-03-19T06:55:29+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fbcc383aa", "title": "PageRank Who? AI Giants Opt for Arbitrary Filters", "url": "https://shital.com/blog/tweets/post/202303190613-pagerank-who-ai-giants-opt-for-arbitrary-filters/", "published_at": "2023-03-19T06:13:37+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fbd02279c", "title": "Our Generation's Moonshots: Like Christmas Every Day", "url": "https://shital.com/blog/tweets/post/202303180249-moonshots-like-christmas/", "published_at": "2023-03-18T02:49:26+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fbd77ec9d", "title": "Khan Academy's GPT-4 Tutor: The Teacher Who Never Gets Tired", "url": "https://shital.com/blog/tweets/post/202303171006-khan-academy-gpt4-tutor/", "published_at": "2023-03-17T10:06:26+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fbe065247", "title": "The Teacher Who Never Sleeps: Khan Academy's GPT-4 Tutor", "url": "https://shital.com/blog/tweets/post/202303171005-the-teacher-who-never-sleeps/", "published_at": "2023-03-17T10:05:07+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fbe723c0c", "title": "Tiny Notion's Big AI Notion: $100M ARR Signals Massive Opportunity", "url": "https://shital.com/blog/tweets/post/202303170641-notion-ai-100m-arr/", "published_at": "2023-03-17T06:41:21+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fbf0bd4d6", "title": "Chinchilla's 2% Solution: Compressing Text with Tiny Models", "url": "https://shital.com/blog/tweets/thread/202303161337-chinchillas-2-percent-solution/", "published_at": "2023-03-16T13:37:43+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fbfab4fa4", "title": "The Chinchilla That GPT-4 Missed", "url": "https://shital.com/blog/tweets/post/202303161043-chinchilla-gpt4-missed/", "published_at": "2023-03-16T10:43:23+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc02009de", "title": "Unveiling GPT-4's Hidden Gems", "url": "https://shital.com/blog/tweets/thread/202303151845-unveiling-gpt4-hidden-gems/", "published_at": "2023-03-15T18:45:06+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc0e0a582", "title": "GPT-4 Fails Better: A Step Towards Robust Reasoning", "url": "https://shital.com/blog/tweets/thread/202303150253-gpt4-fails-better/", "published_at": "2023-03-15T02:53:01+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc10bb9b2", "title": "LLaMA Express: From Zero to ChatGPT in 3 Hours", "url": "https://shital.com/blog/tweets/post/202303140834-llama-express-zero-to-chatgpt/", "published_at": "2023-03-14T08:34:12+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc13660ff", "title": "LLaMA Sprints on M1 Pro: 16 Tokens/s via 4-bit C++ Magic", "url": "https://shital.com/blog/tweets/thread/202303110945-llama-m1-pro-4bit-cpp/", "published_at": "2023-03-11T09:45:56+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc1889b54", "title": "Bing Chat Outsmarts Fake Brands", "url": "https://shital.com/blog/tweets/post/202303100642-bing-chat-outsmarts-fake-brands/", "published_at": "2023-03-10T06:42:08+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc216d0b1", "title": "Bing Chat Outsmarts Amazon's Fake Brands", "url": "https://shital.com/blog/tweets/post/202303100639-bing-chat-outsmarts-fakes/", "published_at": "2023-03-10T06:39:56+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc2a3ac52", "title": "From \"Why\" to \"How Much\": Quantify Your Curiosity", "url": "https://shital.com/blog/tweets/post/202303091516-from-why-to-how-much/", "published_at": "2023-03-09T15:16:46+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc3905f19", "title": "Compressing Life: 18 Years in a 3TB File", "url": "https://shital.com/blog/tweets/thread/202303061552-compressing-life-18-years-in-3tb/", "published_at": "2023-03-06T15:52:04+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc3f42a23", "title": "The Quarky Secret of Original ChatGPT", "url": "https://shital.com/blog/tweets/thread/202303051711-quarky-secret-chatgpt/", "published_at": "2023-03-05T17:11:52+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc43fa6a5", "title": "Model in a Tweet: 188 Characters to 80% ImageNet", "url": "https://shital.com/blog/tweets/thread/202303041335-188-char-imagenet-model/", "published_at": "2023-03-04T13:35:22+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc4ae0845", "title": "Poe-etry in Motion: Quick Access to ChatGPT and Claude", "url": "https://shital.com/blog/tweets/post/202303040854-poe-app-chatgpt-claude/", "published_at": "2023-03-04T08:54:03+00:00" }, { "id": "01a087df-e7b2-706b-8e6c-929fc4dce9d8", "title": "Chinchillas and Flashy Attention: How ChatGPT Got 10X Faster", "url": "https://shital.com/blog/tweets/post/202303020843-chatgpt-10x-faster-chinchillas-flashattention/", "published_at": "2023-03-02T08:43:54+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c270cdd93fe", "title": "New Bing Outsmarts Paid ChatGPT in Math Puzzles", "url": "https://shital.com/blog/tweets/post/202303010135-bing-outsmarts-chatgpt-math-puzzles/", "published_at": "2023-03-01T01:35:54+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c270d352111", "title": "Foundation Models: OpenAI Still King in Lucrative Platform Game", "url": "https://shital.com/blog/tweets/post/202302281102-foundation-models-openai-still-king/", "published_at": "2023-02-28T11:02:51+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c270d4c2a4b", "title": "Bewildered by Polls? Blame Evolution!", "url": "https://shital.com/blog/tweets/thread/202302261703-bewildered-by-polls-blame-evolution/", "published_at": "2023-02-26T17:03:10+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c270e0148e7", "title": "No ProbLLaMA: GPT-3 Rival Runs on Single GPU", "url": "https://shital.com/blog/tweets/thread/202302250649-no-problama-gpt3-rival-single-gpu/", "published_at": "2023-02-25T06:49:04+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c270ec26ae2", "title": "Leaps of Faith: The Secret Behind Great Papers", "url": "https://shital.com/blog/tweets/post/202302221432-leaps-of-faith/", "published_at": "2023-02-22T14:32:48+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c270f6e9ebb", "title": "When AI Can't Speak Asimov: The 'Last Question' Test", "url": "https://shital.com/blog/tweets/post/202302161155-ai-cant-speak-asimov/", "published_at": "2023-02-16T11:55:07+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c270f6f3c9b", "title": "Love at First Push: Releasing the Repo on Valentine's Day", "url": "https://shital.com/blog/tweets/post/202302161044-love-at-first-push/", "published_at": "2023-02-16T10:44:29+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2710434396", "title": "From R to PG: AI's Next Billion-Dollar Idea", "url": "https://shital.com/blog/tweets/post/202302151656-from-r-to-pg-ai-idea/", "published_at": "2023-02-15T16:56:01+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2710a08d74", "title": "Bias Cut: Tailoring Transformers for Faster Training", "url": "https://shital.com/blog/tweets/post/202302151240-bias-cut-transformer-training/", "published_at": "2023-02-15T12:40:41+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c27116b2f9f", "title": "Defeating the Hyperparameter Hydra: A Breakthrough in Deep Learning", "url": "https://shital.com/blog/tweets/thread/202302151118-defeating-hyperparameter-hydra/", "published_at": "2023-02-15T11:18:45+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c27118cbc25", "title": "Knowledge Graph Heist: Extracting LLMs by Asking Nicely", "url": "https://shital.com/blog/tweets/post/202302101549-knowledge-graph-heist-llms/", "published_at": "2023-02-10T15:49:09+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2711cc8ed0", "title": "Teaching Programming Without Computers: Knuth Would Be Proud", "url": "https://shital.com/blog/tweets/post/202302081729-teaching-programming-without-computers/", "published_at": "2023-02-08T17:29:00+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2712a47d55", "title": "Vocab Sizes: From Guilty Tens to Speedy Sixty-Fours", "url": "https://shital.com/blog/tweets/post/202302060939-vocab-sizes-tens-to-sixty-fours/", "published_at": "2023-02-06T09:39:52+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2713619810", "title": "Small But Mighty: 3B Model Outsmarts 58X Larger LLM", "url": "https://shital.com/blog/tweets/post/202302031605-small-but-mighty-3b-model/", "published_at": "2023-02-03T16:05:23+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c27140012f1", "title": "Small Data, Big Gains: Pre-training Revolution", "url": "https://shital.com/blog/tweets/post/202302031525-small-data-big-gains-pre-training-revolution/", "published_at": "2023-02-03T15:25:19+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2714696f39", "title": "A Byte-Sized Guide to Mind-Blowing Data Scales", "url": "https://shital.com/blog/tweets/post/202301241946-byte-sized-data-scales/", "published_at": "2023-01-24T19:46:37+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c271538cae0", "title": "Transformers: Bad Band-Aids and Missing Memory", "url": "https://shital.com/blog/tweets/post/202301231611-transformers-bad-band-aids-missing-memory/", "published_at": "2023-01-23T16:11:49+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2716303ecc", "title": "AGI's Final Exam: Peace in the Middle East", "url": "https://shital.com/blog/tweets/post/202301231310-agi-final-exam-middle-east/", "published_at": "2023-01-23T13:10:39+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c271713ea4a", "title": "The $4B Barrier to Beating Google", "url": "https://shital.com/blog/tweets/thread/202301231233-4b-barrier-beating-google/", "published_at": "2023-01-23T12:33:14+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c27173b6541", "title": "Don't Battle the Old Model—Obsolete It!", "url": "https://shital.com/blog/tweets/post/202301231141-dont-battle-old-model-obsolete-it/", "published_at": "2023-01-23T11:41:38+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2717e678b6", "title": "When Docs Talk Back: Q&A with GPT-3 and Embeddings", "url": "https://shital.com/blog/tweets/post/202301231129-when-docs-talk-back/", "published_at": "2023-01-23T11:29:30+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2717ebc2ff", "title": "Dark Mode for Writing, Light Mode for Righting", "url": "https://shital.com/blog/tweets/post/202301230738-dark-mode-writing-light-mode-righting/", "published_at": "2023-01-23T07:38:47+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c27181291b6", "title": "Lost in Hyperspace: The Quest for Mathematical Proofs", "url": "https://shital.com/blog/tweets/post/202301230706-lost-in-hyperspace-quest-for-proofs/", "published_at": "2023-01-23T07:06:20+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2719058d8a", "title": "Compute or Compete: Data Skills Lead the Way", "url": "https://shital.com/blog/tweets/post/202301220741-compute_or_compete/", "published_at": "2023-01-22T07:41:03+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2719b3257b", "title": "LLM Inference on a Shoestring Budget", "url": "https://shital.com/blog/tweets/post/202301211053-llm-inference-shoestring-budget/", "published_at": "2023-01-21T10:53:37+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2719bfeb29", "title": "Run, ZIP, Run: Python Executes ZIP Files", "url": "https://shital.com/blog/tweets/post/202301210827-python-executes-zip-files/", "published_at": "2023-01-21T08:27:29+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c271ab3ba34", "title": "ChatGPT with Citations: Farewell to Blue Links", "url": "https://shital.com/blog/tweets/post/202301201551-chatgpt-with-citations/", "published_at": "2023-01-20T15:51:34+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c271b9e89ba", "title": "GPU Idle? PyTorch's Holistic Trace Analysis Knows Why", "url": "https://shital.com/blog/tweets/post/202301200916-gpu-idle-pytorch-hta/", "published_at": "2023-01-20T09:16:54+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c271c431a95", "title": "Dude, Where's My BERT LLM? Answering Your LLM FAQs", "url": "https://shital.com/blog/tweets/thread/202301181732-dude-wheres-my-bert-llm/", "published_at": "2023-01-18T17:32:47+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c271cac8ac9", "title": "AI Startups Evolve: The New Cambrian Explosion", "url": "https://shital.com/blog/tweets/thread/202301131702-ai-startups-cambrian-explosion/", "published_at": "2023-01-13T17:02:55+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c271d7a828d", "title": "When 4,000 GPUs Feel Like an Abacus", "url": "https://shital.com/blog/tweets/post/202301110931-gpu-abacus/", "published_at": "2023-01-11T09:31:23+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c271dc2aa84", "title": "Copernicus Redux: Not the Center of Creativity", "url": "https://shital.com/blog/tweets/post/202301101933-copernicus-redux-not-center/", "published_at": "2023-01-10T19:33:26+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c271ebcbcee", "title": "Quality Over Speed: My ChatGPT Insight", "url": "https://shital.com/blog/tweets/thread/202301090837-quality-over-speed-chatgpt/", "published_at": "2023-01-09T08:37:50+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c271fb53c0c", "title": "Embedding Magic: ArXiv Search Outsmarts Google", "url": "https://shital.com/blog/tweets/post/202301090619-embedding-magic-arxiv-search/", "published_at": "2023-01-09T06:19:47+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c272066bd73", "title": "VALL-E Wows with 3-Second Voice Cloning", "url": "https://shital.com/blog/tweets/post/202301061543-vall-e-voice-cloning/", "published_at": "2023-01-06T15:43:21+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c27210b34cf", "title": "Too Much Science, Not Enough Disruption?", "url": "https://shital.com/blog/tweets/thread/202301061246-too-much-science-not-enough-disruption/", "published_at": "2023-01-06T12:46:28+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2721d06f49", "title": "Meet DALL-E: Where WALL-E Meets Salvador Dali", "url": "https://shital.com/blog/tweets/post/202301061131-meet-dall-e-wall-e-meets-dali/", "published_at": "2023-01-06T11:31:49+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2722280acd", "title": "U-Net's 3D Time Travel: Filling in Occluded Frames", "url": "https://shital.com/blog/tweets/post/202301060925-u-net-3d-time-travel/", "published_at": "2023-01-06T09:25:12+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2722ca0cab", "title": "Analysts Scramble; I Sense Normalcy Returning", "url": "https://shital.com/blog/tweets/post/202212311631-analysts-scramble-normalcy-returning/", "published_at": "2022-12-31T16:31:47+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c272389f6e8", "title": "Zipf Happens: Transformers Love Uneven Data", "url": "https://shital.com/blog/tweets/post/202212311555-zipf-happens-transformers-love-uneven-data/", "published_at": "2022-12-31T15:55:29+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2723af976f", "title": "Old but Gold: Basic Transformers Still Best at Scaling", "url": "https://shital.com/blog/tweets/post/202212311533-basic-transformers-best-at-scaling/", "published_at": "2022-12-31T15:33:28+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c272401011f", "title": "CrossClimate 2: Snow Problem!", "url": "https://shital.com/blog/tweets/post/202212241137-crossclimate2-snow-problem/", "published_at": "2022-12-24T11:37:09+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2724957c0d", "title": "Get in the Fourier Spirit with Physics Elliot", "url": "https://shital.com/blog/tweets/post/202212231314-fourier-spirit-elliot/", "published_at": "2022-12-23T13:14:21+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c27258b8967", "title": "Woodn't You Know: 3D Printing Wood Without Trees", "url": "https://shital.com/blog/tweets/post/202212200631-woodnt-you-know-3d-printing-wood/", "published_at": "2022-12-20T06:31:54+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2725ff1e3c", "title": "Cutting the Fluff: Time-Saving Features Ahead", "url": "https://shital.com/blog/tweets/post/202212190538-cutting-the-fluff/", "published_at": "2022-12-19T05:38:17+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c27266b1a12", "title": "Norvig vs AlphaCode: Code Quality Showdown", "url": "https://shital.com/blog/tweets/post/202212170510-norvig-vs-alphacode-showdown/", "published_at": "2022-12-17T05:10:46+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2726df15fe", "title": "Queries in English? My 300-Line SQL Is Crying", "url": "https://shital.com/blog/tweets/post/202212161423-queries-in-english-300-line-sql/", "published_at": "2022-12-16T14:23:18+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c27276619cb", "title": "Greed: Humanity's AI Bootloader", "url": "https://shital.com/blog/tweets/post/202212160755-greed-humanitys-ai-bootloader/", "published_at": "2022-12-16T07:55:57+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c2727dbfe9f", "title": "Revenge of RL: How Instruct Paradigm Boosts ChatGPT", "url": "https://shital.com/blog/tweets/post/202212150647-revenge-of-rl-chatgpt/", "published_at": "2022-12-15T06:47:04+00:00" }, { "id": "01a087df-e7b3-728b-ad3f-9c272813c082", "title": "Revenge of RL: The Cherry on ChatGPT's Cake", "url": "https://shital.com/blog/tweets/post/202212150639-revenge-of-rl-chatgpts-cherry/", "published_at": "2022-12-15T06:39:17+00:00" } ] posts Claim your blog

Back to shital.com

Blog · corpus.blog/blogs/shital.com/posts

shital.com

shital.com

2023

Neural Nets Shrink the Planet: Introducing LGSMs original ↗

7 Jun 2023

The 26 Million Reasons Apple Thinks Vision Pro Will Soar original ↗

7 Jun 2023

Leaderboard Lies: Broken Metrics from For-Hire Raters original ↗

26 May 2023

Model Fails Hinton Test: GPT-4 Still Undefeated original ↗

25 May 2023

AI Bots Outsmart Google Rankings original ↗

25 May 2023

Knuth's Verdict on GPT-4: Only 0.01% to AGI original ↗

21 May 2023

PaLM2-L's Token Diet: Outsmarting Chinchilla and LLaMA original ↗

18 May 2023

When Gödel Crashes the AI Alignment Party original ↗

14 May 2023

When Yellow Fades: GPT-4 Passes Hinton's Paint Test While Bard Falls Short original ↗

12 May 2023

GPT-4 Passes Hinton's Paint Puzzle; Bard Falls Short original ↗

12 May 2023

6B Params and a Dream: GPT-JT Challenges GPT-3 original ↗

4 May 2023

Code First, Ask Questions Later: The Understudied Secret Behind GPT-3.5's Success original ↗

1 May 2023

LLaMA Distilled: How GPT-4 Fuels Fine-Tuning Ingenuity original ↗

30 Apr 2023

PyTorch Smuggles CUDA Onboard original ↗

25 Apr 2023

Big Flops, Bigger Bills: GPT-4's $43M Training Cost original ↗

23 Apr 2023

Chill Out Your LLM: Quick Temperature Tuning for Calibration original ↗

12 Apr 2023

FlashAttention's Oscar-Worthy Performance original ↗

8 Apr 2023

FlashAttention: AI's Oscar-Worthy Money Saver original ↗

8 Apr 2023

Segment Anything vs. The Torture Test: Zero-Shot Showdown original ↗

6 Apr 2023

AMD MI250x: The Overlooked Powerhouse Outsmarting A100 original ↗

29 Mar 2023

Midjourney v5: Hands Up, Letters Down original ↗

27 Mar 2023

Beam Me Up, Decoder!original ↗

27 Mar 2023

Be the Prompt, Not the Typist: GPT-4’s Token Triumph original ↗

26 Mar 2023

Transformers: More Than Meets the BLEU Score original ↗

25 Mar 2023

Casually Dropping GPT-4's Trillion Parameters Bombshell original ↗

25 Mar 2023

Bob Metcalfe Wins 2023 Turing Award, Remains Brutally Honest original ↗

23 Mar 2023

Bard's Brain Drain: Tiny Model, Can't Code, Needs Search original ↗

22 Mar 2023

Bard's Coding Conundrum: Smaller Model, Bigger Prompts original ↗

22 Mar 2023

Humans: God's Alignment Problem?original ↗

21 Mar 2023

Say Goodbye to the 'L's in LLMs original ↗

19 Mar 2023

The OPT Disasterpiece: LLM Training Gone Wrong original ↗

19 Mar 2023

PageRank Who? AI Giants Opt for Arbitrary Filters original ↗

19 Mar 2023

Our Generation's Moonshots: Like Christmas Every Day original ↗

18 Mar 2023

Khan Academy's GPT-4 Tutor: The Teacher Who Never Gets Tired original ↗

17 Mar 2023

The Teacher Who Never Sleeps: Khan Academy's GPT-4 Tutor original ↗

17 Mar 2023

Tiny Notion's Big AI Notion: $100M ARR Signals Massive Opportunity original ↗

17 Mar 2023

Chinchilla's 2% Solution: Compressing Text with Tiny Models original ↗

16 Mar 2023

The Chinchilla That GPT-4 Missed original ↗

16 Mar 2023

Unveiling GPT-4's Hidden Gems original ↗

15 Mar 2023

GPT-4 Fails Better: A Step Towards Robust Reasoning original ↗

15 Mar 2023

LLaMA Express: From Zero to ChatGPT in 3 Hours original ↗

14 Mar 2023

LLaMA Sprints on M1 Pro: 16 Tokens/s via 4-bit C++ Magic original ↗

11 Mar 2023

Bing Chat Outsmarts Fake Brands original ↗

10 Mar 2023

Bing Chat Outsmarts Amazon's Fake Brands original ↗

10 Mar 2023

From "Why" to "How Much": Quantify Your Curiosity original ↗

9 Mar 2023

Compressing Life: 18 Years in a 3TB File original ↗

6 Mar 2023

The Quarky Secret of Original ChatGPT original ↗

5 Mar 2023

Model in a Tweet: 188 Characters to 80% ImageNet original ↗

4 Mar 2023

Poe-etry in Motion: Quick Access to ChatGPT and Claude original ↗

4 Mar 2023

Chinchillas and Flashy Attention: How ChatGPT Got 10X Faster original ↗

2 Mar 2023

New Bing Outsmarts Paid ChatGPT in Math Puzzles original ↗

1 Mar 2023

Foundation Models: OpenAI Still King in Lucrative Platform Game original ↗

28 Feb 2023

Bewildered by Polls? Blame Evolution!original ↗

26 Feb 2023

No ProbLLaMA: GPT-3 Rival Runs on Single GPU original ↗

25 Feb 2023

Leaps of Faith: The Secret Behind Great Papers original ↗

22 Feb 2023

When AI Can't Speak Asimov: The 'Last Question' Test original ↗

16 Feb 2023

Love at First Push: Releasing the Repo on Valentine's Day original ↗

16 Feb 2023

From R to PG: AI's Next Billion-Dollar Idea original ↗

15 Feb 2023

Bias Cut: Tailoring Transformers for Faster Training original ↗

15 Feb 2023

Defeating the Hyperparameter Hydra: A Breakthrough in Deep Learning original ↗

15 Feb 2023

Knowledge Graph Heist: Extracting LLMs by Asking Nicely original ↗

10 Feb 2023

Teaching Programming Without Computers: Knuth Would Be Proud original ↗

8 Feb 2023

Vocab Sizes: From Guilty Tens to Speedy Sixty-Fours original ↗

6 Feb 2023

Small But Mighty: 3B Model Outsmarts 58X Larger LLM original ↗

3 Feb 2023

Small Data, Big Gains: Pre-training Revolution original ↗

3 Feb 2023

A Byte-Sized Guide to Mind-Blowing Data Scales original ↗

24 Jan 2023

Transformers: Bad Band-Aids and Missing Memory original ↗

23 Jan 2023

AGI's Final Exam: Peace in the Middle East original ↗

23 Jan 2023

The $4B Barrier to Beating Google original ↗

23 Jan 2023

Don't Battle the Old Model—Obsolete It!original ↗

23 Jan 2023

When Docs Talk Back: Q&A with GPT-3 and Embeddings original ↗

23 Jan 2023

Dark Mode for Writing, Light Mode for Righting original ↗

23 Jan 2023

Lost in Hyperspace: The Quest for Mathematical Proofs original ↗

23 Jan 2023

Compute or Compete: Data Skills Lead the Way original ↗

22 Jan 2023

LLM Inference on a Shoestring Budget original ↗

21 Jan 2023

Run, ZIP, Run: Python Executes ZIP Files original ↗

21 Jan 2023

ChatGPT with Citations: Farewell to Blue Links original ↗

20 Jan 2023

GPU Idle? PyTorch's Holistic Trace Analysis Knows Why original ↗

20 Jan 2023

Dude, Where's My BERT LLM? Answering Your LLM FAQs original ↗

18 Jan 2023

AI Startups Evolve: The New Cambrian Explosion original ↗

13 Jan 2023

When 4,000 GPUs Feel Like an Abacus original ↗

11 Jan 2023

Copernicus Redux: Not the Center of Creativity original ↗

10 Jan 2023

Quality Over Speed: My ChatGPT Insight original ↗

9 Jan 2023

Embedding Magic: ArXiv Search Outsmarts Google original ↗

9 Jan 2023

VALL-E Wows with 3-Second Voice Cloning original ↗

6 Jan 2023

Too Much Science, Not Enough Disruption?original ↗

6 Jan 2023

Meet DALL-E: Where WALL-E Meets Salvador Dali original ↗

6 Jan 2023

U-Net's 3D Time Travel: Filling in Occluded Frames original ↗

6 Jan 2023

2022

Analysts Scramble; I Sense Normalcy Returning original ↗

31 Dec 2022

Zipf Happens: Transformers Love Uneven Data original ↗

31 Dec 2022

Old but Gold: Basic Transformers Still Best at Scaling original ↗

31 Dec 2022

CrossClimate 2: Snow Problem!original ↗

24 Dec 2022

Get in the Fourier Spirit with Physics Elliot original ↗

23 Dec 2022

Woodn't You Know: 3D Printing Wood Without Trees original ↗

20 Dec 2022

Cutting the Fluff: Time-Saving Features Ahead original ↗

19 Dec 2022

Norvig vs AlphaCode: Code Quality Showdown original ↗

17 Dec 2022

Queries in English? My 300-Line SQL Is Crying original ↗

16 Dec 2022

Greed: Humanity's AI Bootloader original ↗

16 Dec 2022

Revenge of RL: How Instruct Paradigm Boosts ChatGPT original ↗

15 Dec 2022

Revenge of RL: The Cherry on ChatGPT's Cake original ↗

15 Dec 2022