DailyPost 3040
DEEPSEEK-R1
If the world is up for grabs, the battle is bound to be fierce, engaging, full of surprises, see-saw many times, any number of competitors, and a roller coaster ride till the final winner is decided. At the end there can be many winners, who knows. It might also happen that a long battle of attrition might ensue, with the end not known. The only known fact is that the world will gain immensely out of it, and human civilization will get elevated to a different level. You might be trying to make a guess, and if you have been following the tech world in recent times, you would for sure guess it right. This is the battle for supremacy in AI, the earth shattering technology.
The news emanating out of the AI stable is much more than all the other breaking news combined globally. AI deserves every single coverage, given the fact that it has completely changed the world in the last two years. When ChatGPT hit the world as the conversational GPT, it took the world by storm like no other technology product in the history of mankind. It was presumed OpenAI would rule the roost for many years to come. Ruling apex technology for long is no easy job. The IT behemoths were not far behind with Gemini and Copilot. The lesser ones also gave a run for the money.
In this fast moving AI landscape, the lesser mortals were Claude (Anthropic), Perplexity AI, Jasper AI and the like were making their presence seriously felt. But if someone thought that lead would remain for long with the US, that may turn out to be a pipedream. China’s rapid growth in this field is undeniable. A Chinese AI lab DeepSeek has recently launched DeepSeek-V3. In a very short time it is back again with yet another powerful reasoning large language model named the DeepSeek-R1 is making heads turn globally. The architecture is known as the mixture of experts. As the benchmark still remains OpenAI, comparison with its frontier model o1 is natural.
DeepSeek-R1 matches the performance in tasks like math, coding and general knowledge. What is really fascinating is that DeepSeek-R1 is reportedly 90-95% more affordable than o1. It reasons through the problems like humans. It is an open-source reasoning model. DeepSeek-V3 earlier outperformed models by Meta and OpenAI while developed at a fraction of their cost. It has been trained entirely via reinforcement learning, without any supervised fine tuning. In short, the new DeepSeek model is a state-of-the-art reasoning model designed to enhance problem solving and analytical capabilities of AI systems.
THE AI TSUNAMI HAS SWEPT US OFF OUR FEET. A NEW WORLD IS BORN.
Sanjay Sahay