Large Language Models (LLMs) like OpenAI’s ChatGPT are going to become more affordable and common because of advancements in new technology. Last week, a Chinese AI firm released their open-source AI model called DeepSeek R1 model which competes with OpenAI’s o1 in both performance and cost. This new model of DeepSeek caused a lot of effect on Nvidia’s stocks, with the company losing almost $600 billion in market value. This is the largest single-day drop in the history of the US. Many investors are worried that the rise of open-source AI models can reduce the demand for Nvidia’s graphics processors which are important for training and running AI systems. Many tech CEOs are also shifting focus from AI models to AI agents which will also lessen the value of traditional LLMs and will make LLMs a commodity.
R1 also stands out among other models because it uses a mixed precision framework, combining full-precision 32-bit floating point (FP32) numbers and low-precision 8-bit floating point (FP8) numbers. It uses FP8 for most calculations and switches to FP32 when more accuracy is needed for specific tasks. Satya Nadella, the CEO of Microsoft, says that AI will turn into a commodity as it becomes more accessible and efficient just like DeepSeek’s new model. Appian CEO also agreed to this statement and also added that as companies develop competitive AI, the lack of differentiation will be bad for high spending first movers.
There’s been a hype about AI models like ChatGPT since the past two years. They are LLM powered apps and products which work upon command and user input. On the other hand, AI agents are more action oriented and work without much user input to complete tasks. For example, if you want to book a doctor’s appointment, you may ask ChatGPT to find you the nearest GP. In future, AI agents will not only be able to find you the nearest GP, they will be able to find the available times and book your appointment.
CEO of Mistral, Arthur Mensch, said that now the focus should be on models which integrate both language models as well as contextual business data. He also added that many companies are adapting to AI and distilling the expertise into those systems. An AI agent called Operator has been released by OpenAI, while Anthropic, which is an AI startup backed by Amazon, has also introduced its own agent called “computer use” last year.
Image: DIW-aigen
Read next:
• Is Your System Secure? Hackers Are Now Attacking 22% Faster Than Last Year
• Google’s AI Overviews Now Handle Every Complex Query – Here’s How It’s Changing Search Forever
• Gamelight Powers Springcome's Merge Sweets Global Success with 500K Installs
• New Survey Shows Goals and Challenges Small Businesses Have and What they Think of AI
Hugging Face co-founder Thomas Wolf envisions a future where LLMs deeply embed into intelligent systems tied to company databases. Because of open-source models, many AI tools are going to be easily accessible and there will be less focus on models themselves. He also compared the shift in AI to the internet revolution where the focus moved from building websites to creating internet native companies. He said that in future, AI is also going to be used for building systems and tools for specific tasks. The rise in DeepSeek R1 also makes tech CEOs believe that LLMs are becoming commoditized.
R1 also stands out among other models because it uses a mixed precision framework, combining full-precision 32-bit floating point (FP32) numbers and low-precision 8-bit floating point (FP8) numbers. It uses FP8 for most calculations and switches to FP32 when more accuracy is needed for specific tasks. Satya Nadella, the CEO of Microsoft, says that AI will turn into a commodity as it becomes more accessible and efficient just like DeepSeek’s new model. Appian CEO also agreed to this statement and also added that as companies develop competitive AI, the lack of differentiation will be bad for high spending first movers.
There’s been a hype about AI models like ChatGPT since the past two years. They are LLM powered apps and products which work upon command and user input. On the other hand, AI agents are more action oriented and work without much user input to complete tasks. For example, if you want to book a doctor’s appointment, you may ask ChatGPT to find you the nearest GP. In future, AI agents will not only be able to find you the nearest GP, they will be able to find the available times and book your appointment.
CEO of Mistral, Arthur Mensch, said that now the focus should be on models which integrate both language models as well as contextual business data. He also added that many companies are adapting to AI and distilling the expertise into those systems. An AI agent called Operator has been released by OpenAI, while Anthropic, which is an AI startup backed by Amazon, has also introduced its own agent called “computer use” last year.
Image: DIW-aigen
Read next:
• Is Your System Secure? Hackers Are Now Attacking 22% Faster Than Last Year
• Google’s AI Overviews Now Handle Every Complex Query – Here’s How It’s Changing Search Forever
• Gamelight Powers Springcome's Merge Sweets Global Success with 500K Installs
• New Survey Shows Goals and Challenges Small Businesses Have and What they Think of AI