News

According to BlockBeats, SoftBank is reportedly in discussions with OpenAI regarding a potential direct investment ranging from $15 billion to $25 billion. This development follows SoftBank's previous commitment to invest over $15 billion in the 'Stargate' project, as reported by the Financial Times on January 30.

According to Odaily, data from CoinGecko indicates that the total market capitalization of AI-related tokens has decreased to $9.594 billion, marking an 8.7% drop over the past 24 hours. Among the tokens, FET experienced a 6.3% decline, with its market value now at $2.542 billion. AI16Z saw a 1.2% decrease, bringing its market capitalization to $695 million. Meanwhile, VIRTUAL's market value fell by 0.1%, currently standing at $1.26 billion.

According to PANews, U.S. President Donald Trump's nominee for Commerce Secretary, Lutnick, emphasized the importance of the United States leading in the field of artificial intelligence and establishing relevant standards. Additionally, Lutnick highlighted the necessity for stablecoin reserves to undergo audits. He also praised the chip manufacturing subsidy program but noted that investment in this area requires careful review.

According to Odaily, JPMorgan has indicated that more affordable artificial intelligence models are expected to stimulate incremental growth. These advancements are anticipated to deliver early efficiency gains and contribute to reducing inflation.

According to Odaily, the U.S. Navy has instructed its personnel to refrain from using AI technology developed by DeepSeek. This directive aims to ensure security and operational integrity within the Navy's operations. The decision reflects ongoing concerns about the potential risks associated with AI technologies in sensitive military environments.

According to Odaily, market reports indicate that Microsoft (MSFT.O) is conducting an investigation into whether organizations associated with DeepSeek have accessed OpenAI's data.

According to Odaily, DeepSeek's official website has reported a significant malicious cyber attack. Experts have indicated that the IP addresses involved in this attack are located in the United States.

According to BlockBeats, billionaire hedge fund manager and Bridgewater Associates founder Ray Dalio has issued a warning about a potential bubble in the U.S. stock market driven by investor enthusiasm for artificial intelligence. Dalio compared the current situation to the dot-com bubble that burst at the turn of the century.Dalio highlighted that the current pricing levels in the U.S. stock market are extremely high and coupled with interest rate risks, this combination could potentially burst the bubble. He noted that the current phase of the market cycle is very similar to the conditions in 1998 or 1999. In his view, while there is no doubt that new major technologies will change the world and achieve success, some investors may be confusing this technological success with investment success.

According to Odaily, OpenAI CEO Sam Altman recently expressed his admiration for DeepSeek's R1 model on X, highlighting its impressive performance, particularly in terms of cost-effectiveness. Altman acknowledged the presence of new competitors in the field, which he finds exciting, and mentioned that OpenAI plans to release new versions of its models. He emphasized the importance of continuing their research roadmap and noted the increasing need for computational power to achieve their mission. Altman anticipates that the next generation of AI models will be remarkable and looks forward to sharing advancements in artificial general intelligence (AGI) and other areas.

According to BlockBeats, U.S. President Donald Trump announced on January 28 that Microsoft is in negotiations to acquire TikTok. The President expressed hope for a competitive bidding process surrounding TikTok.

According to Odaily, Nvidia's stock experienced a significant drop of 17% at the close of the U.S. stock market on January 27. In response, Nvidia highlighted the advancements in artificial intelligence through its DeepSeek project. The company described DeepSeek as an exemplary model of testing-time scaling, showcasing how the technology can be utilized to create new models using widely available frameworks and computing power that complies with export regulations. Nvidia emphasized that the inference process requires a substantial number of Nvidia GPUs and high-performance networks. The company also introduced three scaling laws: the enduring pre-training and post-training laws, along with the new testing-time scaling law.

According to BlockBeats, on January 28, the artificial intelligence community Hugging Face announced the release of DeepSeek's open-source multimodal AI model, Janus-Pro. The Janus-Pro-7B model has outperformed OpenAI's DALL-E 3 and Stable Diffusion in GenEval and DPG-Bench benchmark tests.

According to Odaily, former ConsenSys global digital asset architect Ejaaz Ahamadeen recently shared insights on X regarding the transformative impact of DeepSeek on AI development timelines. Ahamadeen noted that DeepSeek has significantly accelerated these timelines, challenging previous assumptions that building such products would require $100 billion and result in a closed-source monopoly. Instead, DeepSeek has surpassed leading models at a minimal cost and has made its technology open source, allowing any AI developer access to this powerful model. Ahamadeen questioned why there is little discussion about the emergence of 'better agents' and suggested that the 'fat application theory' is becoming more relevant. He emphasized that this development is a significant boon for open-source AI.

According to Odaily, ai16z co-founder Shaw recently shared insights on X, emphasizing that more powerful models are always beneficial for AI agents. Over the years, AI labs have consistently outperformed each other in benchmarks and capabilities, with Google, OpenAI, and Claude taking turns in the lead. Currently, DeepSeek is at the forefront. The trend shows that the world's largest and most well-funded companies are competing in a technology that is ultimately becoming free, open-source, and operable on home computers without cost. The consistent winners in this race have been hardware and consumer products. Nvidia remains a constant victor, with each model optimized for its hardware. Apple also benefits, having invested in a unified memory architecture that allows high VRAM machines to run the latest models, albeit slowly. Products continue to gain from the latest models, with examples like Cursor and Perplexity improving significantly every few months. As AI integrates into nearly all products, they all benefit from cheaper and faster AI models. AI agents represent a new application paradigm, with the core argument being that applications need to migrate to social media platforms where users are present. Agents can exist entirely on social media without requiring users to leave, benefiting from network effects with each user interaction. Integrating a new model into an agent framework typically requires only a few lines of code, as most model providers follow the same API conventions established by OpenAI, allowing this integration to be completed in minutes. This enables any agent application to access the latest models immediately, becoming smarter with each new state-of-the-art model. This week marks a significant victory for everyone involved, including agents, humans, and AI model teams, who are now motivated to work harder and improve. Shaw expressed confidence in their position, stating that they are building the next version of Eliza, which will only get better. Thousands of teams are building on their technology, with over 500 contributors to the core repository, and this number is expected to grow as they continue to develop. They are creating a template for ambitious founders to crowdfund their public product projects, with more content to be launched in the coming weeks and months to solidify this strategy.

According to PANews, Tether CEO Paolo Ardoino recently stated on Twitter that future advancements in AI model training will no longer require millions of GPUs. He criticized the current reliance on brute computational power as inefficient and inelegant. Ardoino suggested that superior models will significantly reduce training costs, emphasizing that data acquisition remains crucial. He also predicted a shift towards local or edge-based inference to avoid wasting billions of dollars on mere computational power.

According to Cointelegraph, artificial intelligence (AI) agents are becoming integral to decentralized finance (DeFi), automating financial systems and trading strategies. However, concerns about their safety and reliability persist as their adoption increases. Mike Cahill, CEO of Douro Labs and contributor to the Pyth Network, emphasized the importance of real-time, high-fidelity data for AI agents to make accurate decisions. He noted that errors or data manipulation could lead to unintended consequences, which can be mitigated by providing ultra-low-latency, first-party price updates.AI agents rely on precise and timely data to function effectively in fast-paced markets, ideally sourced directly from first-party providers like market makers or exchanges. These systems aggregate data from multiple sources, reducing the risk of manipulation or inaccuracies, with updates occurring as quickly as one millisecond. This ensures that AI agents can respond instantly to market changes. Cahill highlighted that Pyth ensures agents operate on the most accurate market data available, eliminating risks associated with outdated or manipulated information. AI agents thrive on speed, precision, and automation.One of the significant challenges for AI agents is operating safely within volatile market conditions, which decentralized systems aim to address. For instance, Oracle Integrity Staking (OIS) requires data publishers to stake capital, aligning their financial incentives with the accuracy of their contributions. This means they risk losing their stakes if they provide faulty or manipulated data. Cahill explained that Pyth’s OIS creates an economic security layer that aligns with first-party price sourcing and weighted aggregation to develop resilient, high-frequency pricing that reflects true market conditions. AI agents can also integrate programmable safeguards, such as confidence intervals and predefined slippage thresholds, preventing them from executing trades under volatile or unreliable conditions.Cahill envisions a future where fully autonomous financial systems operate more efficiently than any human-run market ever could, anticipating the emergence of artificial general intelligence (AGI) agents within the next one to three years. He stated that real-time data empowers AI agents to unlock a new era of high-frequency, algorithmic trading in DeFi, where institutional DeFi surpasses traditional finance by offering a market that is not only decentralized but also faster, more efficient, and truly autonomous. Recent developments in the crypto industry, such as Fetch.ai's launch of a $10 million accelerator for AI agent startups and the publication of a white paper by the developer behind ai16z on Web3-native AI agents, underscore the growing push for AI agent technology.

According to Odaily, OpenAI has introduced its first AI agent tool named 'Operator' on Thursday. This tool is designed to assist users in performing tasks based on web operations.

According to PANews, U.S. President Donald Trump stated that to achieve the desired scale of artificial intelligence, the United States needs to double its current energy supply. He emphasized that the energy emergency declaration aims to expedite the regulatory approval process in the country. Trump announced plans to allow rapid construction of power plants that can utilize any necessary fuel. Additionally, he proposed reducing the corporate tax rate to 15% for companies manufacturing products within the United States.

According to PANews, OpenAI's Chief Product Officer Kevin Weil announced that the company plans to release a more advanced GPT-3 model in February or March. Additionally, OpenAI aims to launch its first AI agent tools in the first quarter, enabling ChatGPT to perform practical tasks on computers. Weil provided an example where ChatGPT could assist his child in filling out and submitting a soccer registration form. Meanwhile, Anthropic CEO Dario Amodei, in an interview on January 21, predicted that AI surpassing human intelligence could emerge within two to three years. Weil suggested that this development might even occur before 2027.