DailyPost 1709
GPT 4 – CREATING A FUTURE UNIMAGINED
In most of the narrative pertaining to the second surge, the basic data is not forthcoming, which has put all the services and planning into jeopardy. The world of tech is moving in different direction. If you don’t use the available technology, say genome sequencing, the peril would come right inside your doors. For those who use tech, sky is also not the limit . GPT-3 was great news last year and, in this age, this is the digital overreach we were waiting for. GPT-3, the largest neural network ever created, revolutionized the AI world. The beta version was released by OpenAI and early adopters started finding out of the world results. The results could make anybody wonderstruck. Web page could be transformed to the corresponding code and it could write customized poetry or songs.
GPT-3 was an untrained tool, brute forcing the internet’s data. The AI was so powerful that it became a meta-learner on its own. The barrier which it broke down, was that it learnt how to learn. Users could communicate in natural language, it could interpret. GPT-3 as the name suggests was third in the series, in three years. To put GPT-3 into some perspective, GPT-3 is 100x bigger than its predecessor. GPT-4 is yet to arrive. After the revolutionary GPT-3, what can be expected out of GPT-4?* GPT-4 as being discussed in an extrapolation / speculation of GPT models in general and GPT-3 in particular. In a recent article Alberto Romero does this for us.
GPT-3 is not just a progression over GPT-2. It has been a qualitative leap; it could do things which the earlier models could not. If this trend were to continue which seems likely, then GPT-4 would not just be bigger but it might end up with new qualitative differences. GPT-3 created a learning pedagogy, then what do we expect from the variant in the waiting, who knows what GPT-4 could bring? It would not be out of place to expect the first neural network capable of true reasoning and understanding. This would be when science fiction is likely to come to life. ”GPT-4 may implement a way of assessing the quality of a given prompt.” Conversely, any system which lacks the ability of self-assessment cannot be called intelligent. A system that’s incapable of self-assessment can’t be called intelligent.
For human beings the parameters of intelligence have remained the same for centuries, for machines it has been changing rapidly in its march towards General Artificial Intelligence. GPT-4 is likely to have a much larger context window. The march of AI for general usage rests much on GPT-4. Romero’s last conclusion is “GPT-4 will avoid the limitations of early transformer architectures. The context window will be larger, allowing the system to perform more complex tasks.” The world is on the verge of humongous transformation, it will happen to those who decide to be the early adopters. For others, they also serve who wait and watch.
GPT-4 IS BRINGING GENERAL ARTIFICIAL INTELLIGENCE CLOSER.
Sanjay Sahay