What Is Beyond The Gen AI Buildout?
I am struggling to imagine near-term use cases
The Gen AI buildout is generating significant Revenue for Nvidia (AI chips) and OpenAI (LLMs). However, Gen AI use cases are not clear. Which companies if any will capture meaningful Gen AI-related revenue remains to be seen.
After playing with Google Gemini (Google’s LLM) for the better part of a month and ChatGPT for a few months I can report that my brain synapses aren’t overloaded with potential use cases. My user experience with both ChatGPT and Gemini frankly has been underwhelming.
Have I used Gemini to facilitate the creation of artwork for this newsletter? Yes. Is it a huge win? No.
Have I used Gemini to convert images to tables and tables to spreadsheets with a few keystrokes? Yes. Sounds impressive but is not so impressive in practice. The more data a table contains, the greater the probability of Gemini making errors. Gemini almost always made errors when translating images to tables and tables to spreadsheets. On some occasions there was a modest productivity benefit. On other occasions there was zero benefit. See my recent video posts on the subject HERE and HERE.
Is ChatGPT useful for summarizing writings? Sometimes Yes. On other occasions the errors were so significant that GPT’s output was completely unusable. If I were an I-Banker creating PPT decks with GPT or Gemini, I would triple check the output.
With respect to OpenAI’s Sora, my fear is that there will be enormous privacy and IP-related concerns. Sora’s multi-modal model has cached tons of data (text, images, audio and video) from the public domain and also potentially proprietary data (I don’t trust OpenAI’s CTO’s answer on the subject). I see Sora use cases primarily on the creative side. Video games for example have deployed machine learning for years to render background visuals such as mountain landscapes and city skylines (see our earlier article HERE).
It is the dawn of the LLM Era. The various LLMs and multi-modal models need to significantly evolve in order to drive significant productivity gains for users.