DailyPost 2386
THE PACE AND DEPTH OF ARITIFICIAL INTELLIGENCE
There are many tools which have made news since the days of Deep Blue beating a world chess champion. These tools have caught imagination of the AI buffs, scientists and technologists in this regard could be denied. It never hit the public domain. OpenAI established in 2015 was an attempt towards a focused to pathbreaking AI tools and it started delivering with its ever-evolving GPTs. DeepMind the pioneer in AI was also into high end research and together these two organizations were tipped to define AI trajectory of the world.
With all the progress achieved even reaching the nascent levels of reaching an all-purpose tool was two decades away. This is the reason why AI was still currency. GPT-3 was released in 2020 and then its 3.5 version, which propelled the creation of ChatGPT, which transformed AI for all times to come. The dialogue model and the conversational intelligence built over GPT-3.5 started delivering results from the world. From the day of its release on 30th Nov 2022, it reached 100 million users in less two months of its launch. It has already become a household name globally, and that it holds the promise of world’s future is being accepted by one and all .
An MIT says that this tool helps in reducing the time by half while producing a higher quality output of over 20 percent. It has become and expert assistant you in all fields all the time. New applications are being discovered everyday. The beauty of this machine is that the creators have no idea of what all it can do and by what precise means. And then came GPT-4 on March 2023, less that three and half months after the release of ChatGPT. The input and output word count expanded from 3,000 to 25,000 words. From a purely text model of ChatGPT, GPT-4 become multimodal; audio, video, images text.
It understands what is depicted in the image. From the grocery inside the fridge, it can suggest what dishes can be made. It can understand and solve image-based problems. It can build a website from scratch after being shown a shoddy drawing on paper. It can solve riddles, some really tricky ones and gives the explanation as well. Its improved reasoning capabilities are proved by its performance against benchmarks. It is showing human level common sense in quite a few cases. Hindsight Neglect is another issue which GPT-4 seems to have gotten over. GPT-4 is 40% more factual over ChatGPT. It is 82% less likely to respond ‘disallowed content.’ The sparks of Artificial General Intelligence is getting visible for the first time.
AI TODAY IS ON A SOLID FOOTING WITH VERY LITTLE OR NO CHANCE TO FAIL. WHAT WE MAKE OUT OF IT, WILL BE OUR DESTINY.
Sanjay Sahay