With the swift rise of Chinese AI startup DeepSeek, leading AI institutions in the U.S. are facing stiff competition. As detailed in a research paper authored by DeepSeek scientists, their model R1 outperforms established AI models like OpenAI’s o1 reasoning model in areas such as math, science, and coding. Remarkably, this is achieved at significantly lower development costs compared to the proprietary AI models.
Yann LeCun, head of AI at Meta, believes that the widespread success and adoption of DeepSeek AI is largely due to its open-source design. In other words, he thinks that open-source models tend to outperform proprietary ones. He emphasized the value of open-source models and suggested that DeepSeek could help revive the original mission of OpenAI, which was focused on promoting open-source AI development.
Reportedly, Mark Zuckerberg, Meta CEO, has gathered teams of engineers in four separate rooms to scrutinize the key factor behind DeepSeek’s rapid advancement in AI technology (as reported by The Information). It’s worth noting that DeepSeek’s R1 AI operates on an open-source V3 model, which was developed and refined with around $6 million in investment.
DeepSeek’s AI is giving Meta sleepless nights
Two engineering teams from four workspaces are going to concentrate on finding out exactly how a Chinese AI company reduced the cost of creating and educating its AI technology. Meta plans to utilize this knowledge to steer the development of its upcoming version of Llama AI. Meanwhile, the rest of the teams will investigate the data that was employed to train DeepSeek’s AI model, which could influence the overhaul of Llama AI.
DeepSeek presents a substantial hurdle to American AI corporations such as OpenAI, Anthropic, and Meta, outperforming sophisticated AI models like Meta’s Llama 3.1, OpenAI’s GPT-40, and Anthropic’s Claude Sonnet 3.5 while charging only a small fraction of their costs.
Based on reports from two sources closely tied to Facebook’s parent company, Meta’s AI Infrastructure Director, Mathew Oldham, allegedly hinted that DeepSeek’s main model might exceed the performance of Meta’s upcoming Llama AI, planned for release in early 2025.
While speaking to The Information, a Meta spokesman indicated:
In our ongoing development, we constantly compare all competing models, a practice we’ve followed since the formation of our Gen Al group. Llama has played a crucial role in building the open-source AI model community, and we are thrilled to strengthen our leadership in this field with the imminent launch of Llama 4.
In other locations, the CEO of Meta has mentioned that the company plans to allocate up to $65 billion for their progress in AI technology, which involves building data centers and hiring new staff. It’s worth noting that Mark Zuckerberg envisions a possible future where mid-level AI engineers might outperform professional software engineers at Meta by 2025.
Read More
- EUR JPY PREDICTION
- DF PREDICTION. DF cryptocurrency
- Kingdom Come: Deliverance 2 console commands – All KCD2 cheat codes and how to use them
- EUR RUB PREDICTION
- TRB PREDICTION. TRB cryptocurrency
- How to Get Radzig Kobyla’s Longsword in Kingdom Come: Deliverance 2
- MIN PREDICTION. MIN cryptocurrency
- POL PREDICTION. POL cryptocurrency
- ICP PREDICTION. ICP cryptocurrency
- ZEN PREDICTION. ZEN cryptocurrency
2025-01-28 12:46