Borrowing from the Dr. Strangelove, I would say that it’s time we stop worrying about when AI will kill all humanity and start loving it 🙂
The industry appears to be suffering from cognitive dissonance. On one hand newfound fears about ChatGPT is replacing the euphoria and on the other, a plethora of AI startups have popped up like mushrooms in monsoon. The corporates are banning the use of ChatGPT or AI tools and wanting to jump on the AI bandwagon because of FOMO.
‘Prompt engineering’ is being touted as the holy grail of using LLMs and going by some posts on social media platforms it appears to be the hottest skill on the market. There are folks who are guarantying income of $10K per week by automating content creation and publishing by using their software and ChatGPT! Those of us who lived through the dotcom boom and bust cycle must recognize this brand of cool aid.
Like what happened with the dotcom and other bubbles, most of the newly minted ‘get-rich-quick’ startups that solely rely on ChatGPT or other similar platforms are going to hit a brick wall and fail. The ones that are founded with some solid business plan and meaningful strategy will stand to gain tremendously.
Here’s my cheat sheet on where the opportunities may lie for various segments of the market for capitalizing on the AI. The AI landscape is changing extremely fast. Think of it as Moore’s law on steroids. While access to compute prowess and petabytes of data will continue to be a strategic advantage for the time being, newer approaches may reduce the barriers to entry as is evident by the Chinchilla paper by DeepMind.
Here are some other considerations to help build your AI strategy and adoption roadmap. The old cliché is still valid – it is a marathon, and not a sprint. So, don’t get sucked into FOMO. Think through the strategy before you pull the trigger on any action plan.
- Natural language will evolve to be the primary user interface. Plan for audio and not just textual interface the way ChatGPT and other LLMs have at the moment.
- Prompt Engineering will become an essential skill for data-scientists or ML engineers to help evaluate various LLMs.
- Private LLMs that are fine-tuned to the user’s content will be a gold-mine. Both for large enterprises and startups looking for new products and offerings.
- While the monolithic models will continue to evolve, smaller models that are fine-tuned to a specific task will become more popular; especially, if they are designed to continue learning in real-time.
- TensorFlow’s library for JS will bring machine learning to gazillions of full-stack developers that will accelerate new products and ideas.
Last piece of advice for those who plan to adopt AI – spend the second half of 2023 to evaluate various AI tools and models with targeted POCs. Use the findings to define an AI transformation strategy that will help you create a differentiation amongst your competitors.
For those planning new products or services – focus on personalized education and healthcare.