Large language models are in the limelight due to the stunning adoption of their poster boy, ChatGPT. Racing to a hundred million users two months after its launch, this has set up an arms race among the BigTechs, with every company readying to launch its own LLM and ChatGPT-killer. Microsoft has a headstart integrating ChatGPT into its products, Google stumbled while launching its own chatbot Bard. Meta has announced LLaMa, which it claims is much more powerful than GPT3. Amazon has one in the works, while Baidu has announced Ernie in China, though it is not yet clear whether the Chinese government will allow that. While all this is incredibly exciting, I believe that the real use of this tech and the value is in the ‘verticals’ that these large horizontal models are spawning. Startups building products on these foundation models are mushrooming, with reports that almost 500 of them have already been funded. Most of these models are open, and therefore startups can quickly build very innovative products on top of them. I think of the horizontals to be akin to the cloud, on which so much technology is built. Perhaps going forward, we will have three LLM providers which will dominate this space, much like Microsoft, Amazon and Google do that today.

a. AI Is Becoming More Conversant. But Will It Get More Honest? – The New York Times (nytimes.com) : Character.ai

b. Typical.me

c. Personal.ai

d. Altered.ai : voice altering

e. Tome.ai : story telling

f. Donotpay.com : robot lawyer!

g. Chatsonic: chatgpt on steroids, goes beyond 2021

h. Synthesia.io : text to video

i. Detangle.ai : legal doc to normal text

j. Runway.ml : next gen content generation

k. Poised.ai : AI coach

l. Playgroundai.com : edit photos

m. Paraphrase.ai

n. Powerpresent.ai : Create presentation

o. Murf : text to speech

Two: AI will not replace you, a person using it could: Journalists, artists, programmers, tax advisors are worried that the work they do can now easily be done by ChatGPT and its ilk. That is not entirely correct. GenAI will definitely impact jobs, but if used well, it will enhance most of them. A programmer can use it to write basic code, before putting her creativity and talent to complete it and make it more elegant. A student can use it to generate some ideas for their dissertation, a dermatologist can theoretically use it to narrow down prognosis possibilities, an artist can use DALL

E2 to create some funky art, which she can then work on, and a journalist can structure his piece using it. People who quickly learn the skill of using ChatGPT will be able to work faster and better than people who ignore it; as people using computers, search, and the Internet did. So, do not fear AI, embrace it and better be the person who uses it to replace someone who does not, rather than the other way around!

Three: The GenAI horizontals are interesting, the verticals will be useful: OpenAI’s GPT3 (and soon, GPT4), and competing offerings from Google, Baidu, etc, are interesting, since they are the core large models which have dazzled us with their human-like capabilities. However, what will be really useful are the ‘verticals’ being built with lightning speed on them. The horizontals are like the ‘clouds’ on which these useful apps are built. ChatGPT in a simplistic sense is a chat vertical on GPT3, and there are many others. Elicit allows us to search academic papers for a research question you might have, Copy.ai and Jasper are great tools for creating ad copy, Copilot generates code, Cohere.ai helps you to generate and organise your thoughts. There are hundreds more, with new ones getting created every day. These will be ones which will help you with improve your jobs, your business, and large parts of your life. Watch out for the verticals