New search is in town

We're back from our 4-day long email break. Which means I missed a ton and a bunch of it made it to today's email. But relax, it's not 'real work'. It's just an email, I'll have more stuff tomorrow and the next day. Take your pick at what's interesting, like circling that leftover platter for the 5th time!!

For those that don't know I work at Zapier. I'm going to be working on some exciting AI šŸ¤ Zapier projects in the new year and am looking for folks who want to get early access to Zapier products, push the boundaries of Zapier and AI, give feedback and create cool demos of what's possible. If it sounds interesting to you, click here.

Let's get to it.

šŸ¤Œ Ben's Picks

Large Language Models Encode Clinical Knowledge. So, basically, large language models (LLMs) are really good at understanding and generating human language. However, using them in the medical and clinical fields is a bit more complicated because mistakes in these areas can have serious consequences. To make sure LLMs are up to snuff in these fields, the authors of this research created a new benchmark called MultiMedQA, which tests how well LLMs can answer medical questions, and a new dataset called HealthSearchQA with medical questions that people have searched for online. They also developed a framework for evaluating LLMs and tested two different models. Flan-PaLM did pretty well, but the authors introduced a new method called instruction prompt tuning to improve its performance even further. The resulting model, Med-PaLM, did better, but still wasn't quite as good as human clinicians. Overall, the research suggests that LLMs have the potential to be useful in the medical and clinical fields, but there's still more work to be done to make them reliable.(link)

Email Triager - An email assistant that automatically drafts email replies behind the scenes. Now all those emails can be answered to get us to the haloed Inbox Zero! All you gotta do is hire a VA to click send on each email, jk. (link)

This Detroit startup aims to shape the future of video creation with AI (Sponsor). Small Michigan startup Waymark is trying to break down barriers to video creation with the help of AI. With Waymark, users just type in basic information, and a team of AI models generates a complete marketing video for them in minutes. It looks like this approach is already picking up steam - Waymark has a few enterprise clients under its belt and has been listed in both Hulu and Roku's creative directories. Check it out here.

šŸ› ļø Cool Tools

  • youChat - The AI search assistant that lives in your search engine. I'm going to test using this exclusively over Google and see how it goes for a few days. I tried it with Neeva but didn't have "I'm never going back to Google" vibes. (link)

  • llm-strategy - PoC package based on LangChainAI and OpenAI's GPT that directly connects Python to LLMs. (link)

  • Lookup - Import your data, ask questions, and get results instantly with the power of AI. (link)

  • Mirage instant 3D asset generation - Use images or prompts to create 3D assets in minutes. (link)

  • Portrait plus - Hugging Face Model card with a dreambooth model trained on a diverse set of close to medium range portraits of people. (link)

  • Trelent - AI Docstrings on demand. (link)

  • Howdoi - Open source project to create a chatbot that can hold a conversation, remember previous comments/questions, and answer all types of queries. (link)

  • GPT-3.5 Table-inator - enter a topic and it will generate a table of information with Wikipedia links. (link)

  • LaViLa (Language augmented Video Language Pretraining) - Repurpose LLMs to be visually conditioned ā€œNarratorsā€, and use them to automatically generate video-language paired data. (link)

  • Auto Photoshop StableDiffusion Plugin - Use the capabilities of Automatic1111 Stable Diffusion in Photoshop without switching between programs. (link)

  • elektrif.ai - An AI-dating assistant. (link)

  • Chatbotkit - Easy way for developers and non-developers alike to build chatbots that can communicate with users in natural language. (link)

  • LexGrep - Search through all of Lex Fridman's podcasts in natural language and get answers. (link)

  • Fuzzy Face - Create AI-generated photos of your pet for free. (link)

  • Promptable - Open source tool for creating, managing, testing and deploying your prompts. (link)

  • ClerkieCLI - a terminal tool that uses an LLM to help debug and troubleshoot coding errors in any programming language. (link)

  • ChatBCG - Generative AI for slides. (link)

  • ListGPT - Save, and share chatgpt conversations in public, private or protected mode. (link)

  • Podmate - sends podcast summaries by email minutes after episodes are released. (link)

  • AI Host for your Live Stream - Let artificial intelligence run your live show. (link)

  • Retune - Build GPT-3 powered apps, without writing any code. (link)

šŸŽ“ Learn

  • A super simple tutorial on "Creating a help bot for Slack" using Chatbotkit (mentioned above). And yes, the tutorial is AI-generated. (link)

  • A no-code tutorial on how you can create 3D assets using Generative AI technology. (link)

šŸ”¬ Research

  • Can large language models reason about medical questions? (link)

  • Do DALL-E and Flamingo understand each other? (link)

  • Flow-guided frame-wise neural representations for videos. (link)

  • OPT-IML: Scaling language model instruction meta learning through the lens of generalisation. (link)

  • Neural Shape Compiler: A unified framework for transforming between text, point cloud, and program. (link)

  • A close look at spatial modeling: From attention to convolution. (link)

  • STRUDEL: Structured dialogue summarisation for dialogue comprehension. (link)

  • OpenOutpaint - local offline javascript and html canvas outpainting for stable diffusion Gradio webUI API. (link)

  • Improving automatic discrimination of logos with similar texts. (link)

  • TextBox 2.0: A Text generation library with pre-trained Language Models. (link)

šŸ‘‹ Too many links?! I created a database for all links mentioned in these emails. Refer 1 friend using this link and I'll send over the link database.

šŸ¤“ Everything else

  • The best little unicorn in Texas: Jasper was winning the AI raceā€”then chatGPT blew up the whole game. (link)

  • What to expect from AI in 2023, from TechCrunch. (link)

  • What comes after ChatGPT? 7 predictions for 2023. (link)

  • MIT Technology Reviewā€™s article, ā€˜Whatā€™s next for AI.ā€™ (link)

  • ChatGPT review: Does OpenAIā€™s text AI beat our editorial team? (link)

  • The New York Times interview with Yejin Choi, ā€˜An A.I. pioneer on what we should really fear.ā€™ (link)

  • A thread on top 10 AI spotlights, from a breathtaking 2022 in rewind. (link)

  • Thread on why AI is the best copilot. (link)

  • AI might one day be able to help doctors detect Alzheimerā€™s Disease in its early stages. Research from Drexel Universityā€™s School of Biomedical Engineering, Science and Health Systems recently demonstrated that OpenAIā€™s GPT-3 program can identify clues from spontaneous speech that are 80% accurate in predicting the early stages of dementia. Note that the paper was submitted in August; we've seen much better embeddings since then! (link)

  • Dreambooth Hackathon on Hugging Face. (link)

  • Y Combinator-backed Poly uses AI to generate art assets. (link)

  • The increasing number of GPU startups in China, the competition between AMD, Intel, and Nvidia, and the implications of China's push for tech self-sufficiency. (link)

  • Fine-tune FLAN-T5 for chat & dialogue summarisation. (link)

šŸ–¼ AI images of the day

šŸ¤— Share Ben's Bites

Send this to 1 AI-curious friend and receive my AI project tracker database!

or copy/paste this link: https://bensbites.beehiiv.com/subscribe?ref=PLACEHOLDER

šŸ‘‹ See ya

ā­ļø How did we do?

How was today's email?

Login or Subscribe to participate in polls.

ā­ļø REAL REVIEWS

Reply

or to participate.