Over the past month, I’ve been particularly intrigued by the buzz surrounding DeepSeek, a Chinese AI startup. Their latest release, the AI model powered by R1 V3, has caught my attention due to its impressive performance. Remarkably, this model has outperformed OpenAI’s o1 reasoning model in terms of cost and various benchmarks, such as coding, mathematics, and science.
Therefore, the announcement led to a significant market adjustment, resulting in NVIDIA losing approximately $600 billion in market worth within 24 hours. This decline pushed NVIDIA down from its position as the third most valuable company globally, behind Microsoft and Apple.
On the contrary, Meta’s principal AI scientist, Yann LeCun, contends that there exists a significant misconception regarding the $1 billion investment in AI. According to this executive, the funds allocated to U.S.-based AI companies were primarily intended for inference, rather than training AI models.
In simpler terms, when AI systems use their learned information to understand and process fresh data, this is known as inference. As per the researcher, it’s expected that the costs associated with these advanced inferences made by AI models may increase over time.
When we equip AI systems with video comprehension, logical thinking, extensive memory, and additional abilities, the cost of making inferences naturally rises. Consequently, the public’s response towards DeepSeek seems quite unfairly critical.
In conversation with Business Insider, Thomas Sohmers, the founder of Positron, echoed Lecun’s thoughts regarding the market’s response towards DeepSeek.
The need for inferences and the associated infrastructure costs will significantly increase at a rapid pace. Anyone who overlooks the reduced training costs of DeepSeek without recognizing that this oversight could lead to an exponential growth in inference demand, cost, and spending, is failing to see the big picture by focusing too much on minor details.
The R1 AI model, which has been developed using $6 billion and reinforcement learning, currently leads Apple’s App Store in terms of free AI app downloads in the US, surpassing ChatGPT. To manage increasing popularity and usage, DeepSeek must now prioritize investing in inferencing capabilities to effectively process user requests.
After DeepSeek’s R1 release, Microsoft CEO Satya Nadellas commended the model as highly impressive. He suggested that advancements in AI from China warrant careful attention. Furthermore, he implied that DeepSeek could benefit Microsoft financially amid investor worries about the company’s AI-related expenses without substantial profit gains. “DeepSeek showcases some genuine breakthroughs,” Nadella continued. “When cryptocurrency prices decrease, computing costs for inference also drop. This means more people can use it, leading to an increase in applications.
Read More
- Solo Leveling Season 3: What You NEED to Know!
- AUCTION PREDICTION. AUCTION cryptocurrency
- Owen Cooper Lands Major Role in Wuthering Heights – What’s Next for the Young Star?
- Pokémon Destined Rivals: Release date, pre-order and what to expect
- `Tokyo Revengers Season 4 Release Date Speculation`
- Pregnant Woman’s Dish Soap Craving Blows Up on TikTok!
- XRP CAD PREDICTION. XRP cryptocurrency
- How to Install & Use All New Mods in Schedule 1
- ETH PREDICTION. ETH cryptocurrency
- Disney Cuts Rachel Zegler’s Screentime Amid Snow White Backlash: What’s Going On?
2025-02-03 16:11