To Infinity and Beyond 🤠🚀

Let the games begin, 

Happy New Year! Look at you, still keeping up with your resolutions. I’m so proud of you. Did you get a chance to read our 2022 AI Wrapped written by our very own Calin? Highly recommend! It’ll give you the highlights of all that transpired in the whirlwind year for AI we now know as 2022.

What do we have for you today?

The intellectually curious: how knowledge and learning will be unlocked and elevated in the generative AI era, featuring broadn.The Doctor will see you now: the intersection of LLMs + Health! Google and Deep Mind teamed up to create Med-PaLM.“And who do you think you are?”: ChatGPT supposedly has a little competition (already).

Change learning for good

We’re calling it now, 2023 is the year of entropy for Generative AI. One massive shift we’re anticipating (excitedly) is in education.

Introduced in the latest broadn blog post, my awesome teammates Renganathan ("Renga") Padmanabhan and Calin Drimbau wrote a killer piece on the state of learning—and where broadn is poised to shatter old notions of knowledge discovery and how to learn.

For starters: learning today sucks

Not my words, theirs. But also…are they wrong?

Between cohort-based courses and self-paced modules, options which aren’t ideal or tailored at the individual-level, modern learning platforms aren’t built to design teachings to cater to the individual. But Gen AI has a shot at changing this for the better! How? Through layers (read the blog post, you’ll love the Shrek reference) and adaptation.

For example, if you were in a room with others (or a self-paced program) learning content marketing the material is set and presented the same exact way. The problem is, the way the content is understood, explored, and retained will be different for a financial wiz, a beautician, a content marketer, and a firefighter. But here comes the vision: generative AI is used to select the right sources of content and presents it to you at your experience and interpretation level—even as you’re enrolled in the same exact content marketing course or program with the others. Sounds fascinating? Well, this is broadn.

Generative AI knowledge startup broadn wants to advance your search for knowledge and personal development. With broadn, you essentially have an AI-enabled learning assistant at your disposal for any topic and query you wish to learn about and know more of; coupled with patience and the intent and context of your personal needs. In other words, personalized variants of courses and answers that look, feel, and interact according to you — in unbiased packaging as it collects and shares all points of views.

What once was only a luxury, one can think of broadn as a private tutor. This can be the age where AI skeptics can ease their concerns and we move towards entrusting that AI will help us learn and grow. We can have confidence that its outputs are dependable, accurate, and trustworthy. The mission of broadn is to cut through the noise and provide you the right learning sources to understand topics thoroughly and build your expertise no matter the subject, no matter the question.

Broadn has treats lined up for us this year. I’m getting that “first day back at school” feeling again <good goosebumps> 🎒

Trust me, I'm a Doctor

On Monday researchers from Google revealed they teamed up with Deep Mind and created Med-PaLM, an LLM (instructed by Flan-PaLM) aligned to the medical domain to generate useful and safe clinical and medical-related answers.

Shared by Google Research Scientists Shek Azizi and Vivek Natarajan on Monday, Med-PaLM is so impressive it demonstrated nearly 68% accuracy on MedQA (US Medical License Exam questions), a 17% increase over prior work. Though still not superior to clinicians (thread here), this generative AI is monumental for the medical field. LLMs like Med-PaLM can spark the embers needed to see AI directly impact the future of medicine and clinical practices from both the practitioner and patient side.

Following continued advancements to Med-PaLM as a tool, medical professionals will be able to access information more quickly and accurately than ever before. For patients and consumers, they will be able to receive beneficial and trusted answers to their health questions in a fraction of the time. Prospects of potential LLM utility in the fields of health and medicine is groundbreaking.

Yes, as the research paper points out, there are limitations to Med-PaLM that were found through human evaluations. This simply emphasizes the requirement for a diverse consort of humans to be involved in developing frameworks and the instruction prompt tuning for medical algorithms. However, this is still a large step in the direction of greater accuracy, transparency, and accessibility in the often confusing, biased, and conflicting field of medicine.

I look forward to the milestones made in the world of Generative Healthcare/Medical AI. All this medical talk reminds me, I need my annual physical *glove snaps* 🩺

The new Buzz Lightyear in Generative AI

Like the viral Toy Story meme with Andy saying “I don’t want to play with you anymore”, there’s a new “toy” in the wild (already?!) claiming to humbly be better than ChatGPT.

Enter stage left ChatSonic, a new chatbot experience, by Y Combinator-backed Writesonic, the AI-powered content and blog creation platform, debuting only days ago. Writesonic Founder & CEO Samanyou Garg remarks that ChatSonic beats out ChatGPT in that:

• it’s integrated with Google Search, creating content with the most up-to-date information (unlike ChatGPT which only goes as far as 2021)

• produces factually accurate results unlike ChatGPT (and some other LLMs) which is often only ~60% correct right now

• responds to voice commands like Siri to create your stunning images and content

• gives you the ability to share, edit, and download conversations

Despite its known limitations, ChatGPT is still incredible and has captivated us all. We knew coming into 2023 that Pandora’s box for generative AI would be released. Who knew it would be this soon?

ChatSonic has come to play and we’ll see how it fares in the landscape. ChatGPT has the incumbent recognition (and OpenAI power) meaning ChatSonic can expect a battle to claim “top spot.”

Who will Andy (translation: we) choose? Western cowboy favorite, Woody (ChatGPT) or the Intergalactic space ranger, Buzz (ChatSonic)? Or a third, mystery player yet to reveal themselves?

Find out next time on Dragonba-… wait, wrong franchise. 🤠🚀🪐🦖🥔

New developments to spam your #random Slack channel 💬

• Masterpiece Studio, the first 3D Generative AI creating 3D models from text ready to be used in games and other 3D applications

• An AI Spreadsheet Assistant by Mehran Jalali

• Speech-to-AI tools are popping up: Magical AI, QuickHelp, to name a few

• China’s rushing to build its own Generative AI tech 

Things to learn when you need a raise

• Introducing, There’s An AI For That, an AI database updated regularly with use cases and platforms sourced from all over the world

Want access to the Generative AI investors database? Share and subscribe to AI In the Middle (below) to receive it! ⬇️

Before you go

Inbox getting full? Never-ending list of emails you need to respond to? That might change one day.

Wishing you a prosperous 2023, you’ll hear from me again on Thursday! 🥂

— Matthew J. Sánchez (@matthewjsanchez)