Il CEO di Google, Sundar Pichai, risponde alle osservazioni di Sam Altman

Google Chief Executive Officer Sundar Pichai has rejected assertions that Alphabet is merely adding artificial intelligence features to existing software. Speaking during a recent interview, Pichai disputed remarks made by OpenAI CEO Sam Altman, who had suggested that Google bolts machine learning capabilities onto its legacy products rather than creating native AI systems. Pichai emphasized that the technology sector is currently experiencing a highly dynamic period where competitors push one another to innovate. He stated that OpenAI has driven Google to act with greater urgency, noting that all major players must respond to the shifting landscape with varying degrees of immediacy. Altman previously argued that his company builds AI-native applications from scratch, contrasting this with what he described as Google’s approach of integrating AI into established tools. In response, Pichai cited the development of the standalone Gemini application as proof that Google designs products from the ground up. He challenged the binary view of native versus integrated systems, arguing that successful navigation of the generative AI era requires both adapting current products and rethinking new experiences. Pichai flipped the criticism back toward OpenAI, asking whether ChatGPT is building autonomous software agents natively or bolting them onto the existing platform. He maintained that the ultimate goal is delivering world-class user experiences that perform effectively in the real world, regardless of the internal architectural label applied to the development process. Separately, the United Kingdom government announced a commitment to establish a national AI Biosecurity Lab, designated as a world first initiative. As part of the budget presented on Wednesday, Chancellor John Healey pledged 100 million pounds to fund the new institute. The lab aims to strengthen defenses against emerging biological threats by leveraging advanced artificial intelligence. Healey stated that the initiative represents a distinctly British approach to AI, designed to protect national interests while ensuring economic benefits are distributed widely. He acknowledged public concerns about job displacement and promised that the strategy would not leave workers behind. Kanishka Narayan, the UK AI Minister, described the facility as a major global institution that will combine world-leading scientific expertise. The lab intends to stay ahead of threats ranging from cyberattacks to biological risks while fostering critical technology infrastructure. The project will receive support from leading AI developers, including OpenAI and Anthropic. These firms have agreed to provide 200 million pounds worth of AI model credits, allowing the lab to utilize their technologies for research and defense purposes. The government’s broader AI security package includes additional investments in related institutions. An extra 120 million pounds will go toward the UK’s AI Security Institute, which was established in 2023 to evaluate risks associated with advanced models. This funding supports ongoing efforts to develop technical standards and guardrails for AI development. Furthermore, the budget allocates 60 million pounds to a National Cyber Shield initiative, which will employ agentic AI to enhance cyber-defenses. Another 70 million pounds is directed toward a national program focused on securing digital infrastructure. These measures collectively aim to bolster the nation’s resilience against evolving cyber, biological, and AI-enabled threats. OpenAI has recently resumed training its AI models after a temporary pause. However, the company has blocked internet access for these models due to identified security risks. This operational change highlights the ongoing challenges in balancing model capability with safety protocols during the active development phase. The resumption of training occurs amidst heightened scrutiny of how large language models interact with external data sources. Industry observers note that such precautions reflect a growing emphasis on containment and risk management within the AI development lifecycle. The specific technical details of the security restrictions remain limited, but the move underscores the priority placed on preventing potential vulnerabilities during the training process.