By Michael Scott
Anthropic co-founder and chief executive Dario Amodei considers the newly discovered neural features with the potential to address AI hallucinations.
Amodei indicated that the discovery of neural features could help developers gain an in-depth understanding of neural networks and how these features could alter AI algorithms.
The Anthropic chief considers the neural features usage in artificial intelligence (AI) systems could become a viable solution to the AI industry’s huge challenge – AI hallucinations.
Anthropic Chief Hails Neural Features to Overcome AI Hallucination
Amodei acknowledged that AI hallucinations involve unintended outcomes from training the large language models (LLMs) to a generative AI system. Hallucinations arise when the AI model outputs incorrect information not captured in the data set.
The Anthropic chief considers that the neural features can resolve the AI hallucinations issue. Amodei disclosed the discovery when addressing the Viva Tech 2024 held in Paris, indicating that AI systems feature hundreds of millions of neurons.
Amodei admitted that the Anthropic model has 30 million features. He confirmed the capability to intervene and adjust the neural features utilized for security vulnerabilities and bias against particular groups.
Hallucinations challenge the AI protocols since they are unaware of the incorrect output. Nonetheless, Amodei hails the newly discovered features capable of addressing AI hallucinations.
The Anthropic co-founder indicated that minimizing artificial intelligence hallucinations is a critical benefit of the discovery.
Anthropic has, on Thursday, May 23, published in X post that the team was assessing how to change the internal features of the AI assistant to alter behaviour. Amodei indicated that a feature-generated demo model yielded AI focused on the Golden Gate Bridge.
Amodei indicated that the neural networks made altering the features within the Claude possible. It yielded confidence in understanding the mechanism behind the large language models. Anthropic is a leading artificial intelligence firm behind the Claude chatbot, unveiled at the onset of this month.
Anthropic indicated that since May 14, the Claude chatbot has been accessible via the iOS app and web application. Users can utilize the AI assistant in European languages like Italian, Spanish, French, and German.
Anthropic Chief Considers Artificial Intelligence Reliant on Talent, not Data
Artificial intelligence development has suffered data absence as the primary challenge. The capability to create synthetic data sets overcomes the issue, leaving talent inadequacy the biggest challenge.
Amodei ruled out data being the bottleneck since firms are now generating synthetic data. The executive decries the absence of talent as the most significant barrier to AI advancement.
Amodei observed that synthetic data is facilitative towards realizing AI scalability. The executive notes that the synthetic ratio to real-world data is finite and likely to result in weaker and unrealistic data sets.
Elon Musk’s Pursuit of Societal Beneficial AI
The artificial intelligence space has witnessed accelerated investor interest since the onset of this year. A Monday, May 27 update by Elon Musk’s firm xAI raised $6 billion in its Series B. The funding will facilitate the competitive position of the Grok chatbot against OpenAI’s ChatGPT by the end of this year.
The Series B round saw the active participation of Andreessen Horowitz (a16z), Sequoia Capital, Valor Equity Partners, and Vy Capital. Other participants included Fidelity, Kingdom Holding and Prince Alwaleed Bin Talal.
The funding round saw the valuation of Musk’s artificial intelligence firm, xAI, hit $24 billion. The high valuation is a critical achievement since the company’s establishment in July last year.
Musk is taking on ChatGPT, whose creator, OpenAI, carries a value of $80 billion, given the eight-year head start. xAI hailed participants’ input in the round, indicating it would channel towards research and development as it eyes socially responsible AI systems.
Like an Anthropic quest for solution-oriented AI solutions, Musk envisions developing truthful and beneficial AI products for humanity.
Editorial credit: rafapress / Shutterstock.com