How GPT/LLMs actually work

Prediction below, open to feedback and criticism: As soon as GPT can learn programming languages through self reinforcement learning, it will become invincible and you will be able to create webapps at negligible cost by just giving it objectives. Coding as a career is likely downhill from here on. Right now, we are in the DeepBlue version of GPT. It has learned from all the code available online just like DeepBlue had all human chess moves till then as its training data. But AlphaZero learned on its own from ground up, and similarly led to AlphaGo later on. Future GPT will learn all programming languages on its own from ground up, and deliver fully functional apps at neglible cost.

Could be large company like Infosys or some startups

Yeah, you heard it right. Whatever these “gurus” and course scammers are shilling in the name of prompt engineering and crores of salary is 💩. Heads up, companies are already working on augmenting prompts using Natural Language Generation techniques which build on top of Generative techniques like GPT and other Transformer based Architectures. This is exactly the same as when Search Engines came out and you had to know the right queries to search for to get the answer, perhaps use flags like “site:wikipedia.com” Do we have anything like a Search Query Engineer? As things advance, the human edge of refining prompts will be automated to the highest degrees. Perhaps, Subject Matter Experts will guide for commercial or enterprise level use cases but apart from that…. NO! You will not become rich by becoming a Prompt Engineer. Perhaps, now is the time to look at the next frontiers of AI and work on the cutting edge, that’s the best way to future-proof.

For me? It was GradCAM was a gamechanger at selling computer vision initiatives internally to the non-technical stakeholders. The gradCAM function computes the importance map by taking the derivative of the reduction layer output for a given class with respect to a convolutional feature map. If you have a 5 layered convolutional neural network, then you can use this against any layer and check the class activation map. The best explanation to give is: "Regions in Red are the areas of the image that the neural network is looking at to make a decision" Well to be honest, regions in red represent the class activations arising from that region but, it would be too much for normie business guys.