Deepseek AI Gains Attention, But No Real Innovation | Llm machine learning tutorial | Hands-on large language models pdf | Openai | Turtles AI

Google DeepMind CEO Demis Hassabis acknowledged the value of Deepseek’s AI model, calling it the best Chinese contribution in the field, but downplayed the enthusiasm around it, stressing the absence of scientific breakthroughs. The General AI (AGI) debate is intensifying, with predictions that it will arrive in the next five years.

Key points:

• Deepseek assessment: The Chinese AI model is of high engineering quality, but without real technological advancements.
• Cost efficiency: Deepseek claims to have reduced training costs, but some experts dispute these claims.
• Technological competition: Google DeepMind claims its Gemini 2.0 Flash models are better performers.
• Proximity to AGI: According to Hassabis, an AI with human cognitive capabilities could be developed within five years.

The AI model developed by Deepseek has been described by Demis Hassabis, CEO of Google DeepMind, as China’s best work in this field so far. However, according to Hassabis, despite the impressive engineering capabilities demonstrated by the company, the media hype generated by its innovation is overblown, as it does not introduce real scientific breakthroughs, but rather applies techniques already established in AI. Deepseek recently published a study that had a strong impact on the markets, claiming that it had developed an AI model at a significantly lower cost than the major players in the field, employing less advanced hardware than that used by large technology companies. These statements triggered mixed reactions, including a wave of selling of technology stocks and a debate about the real need for infrastructure resources to develop advanced AI models. According to some experts, however, Deepseek’s stated costs are underestimated compared to reality. Meanwhile, Google DeepMind has launched Gemini 2.0 Flash models, claiming that these are more efficient than the one proposed by Deepseek, thus confirming the continued competition in the field of advanced AI. At the same time, the debate on general AI (AGI) is becoming increasingly heated, with Hassabis believing that the field is rapidly approaching a system that can match human cognitive capabilities. His prediction places this realization within a time frame of about five years, a timeline also reflected in recent statements by Sam Altman, CEO of OpenAI, who expressed confidence in the possibility of building an AGI according to the traditional definition of the term. However, the prospect of such an advanced AI raises significant questions about security and human control, with leading experts such as Max Tegmark and Yoshua Bengio warning of the risks associated with the loss of oversight over autonomous systems.

The evolution of AI continues to redefine technological and geopolitical balances, prompting companies and institutions to question how to manage the benefits and criticalities of this transformation.