מעבר לאינטליגנציה האנושית: קלוד 3.0 והחיפוש אחר AGI

Last week, Anthropic unveiled version 3.0 of their Claude chatbot family. This model follows Claude 2.0 (released just eight months ago) and showcases how fast this industry is evolving.

With this latest edition, Anthropic sets a new standard in artificial intelligence, promising improved capabilities and safety that redefine the competitive landscape dominated by GPT-4 – at least for now. This is another step towards the adaptation or deviation of human-level intelligence, representing progress towards Artificial General Intelligence (AGI). It further highlights questions about the nature of intelligence, the need for ethics in AI, and the future relationship between humans and machines.

Instead of a grand event, Anthropic quietly launched 3.0 in a blog post and in interviews with The New York Times, Forbes, and CNBC. The stories received factual reporting, mainly without the usual hype common with recent AI product launches.

However, the launch was not entirely clean of bold statements. The company said that the top "Opus" model of the line "exhibits nearly human levels of comprehension and fluency in complex tasks, pushing the boundaries of general intelligence" and "shows us the outer limits of what is possible with generative AI." This seems reminiscent of Microsoft's paper from a year ago stating that ChatGPT showed "hints of Artificial General Intelligence."

Like competitive proposals, Claude 3 is multi-model, meaning it can respond to text and image queries, such as image analysis or diagram interpretation. Currently, Claude does not produce images from text, and perhaps this is a wise decision based on the current challenges related to this ability. Claude's features are not only competitive but – in some cases – leading in the industry.

There are three versions of Claude 3, ranging from the entry-level "Haiku" to the nearly expert "Sont" and the flagship "Opus." All of them include a context window of 200,000 tokens, equivalent to about 150,000 words. This extended context window allows models to analyze and answer questions about large documents, including research articles and novels. Claude 3 also offers leading results in standard language and math tests, as seen below.

Any doubts about Anthropic's ability to compete with market leaders were dispelled with this launch, at least for now.

According to AI expert Gary Marcus, AGI requires systems that can understand and learn from their environments in a general way, with self-awareness and the ability to apply thinking across diverse domains. While recent advancements in LLM like Claude 3 excel in specific tasks, AGI requires a level of flexibility, adaptability, and understanding that it and other current models have not yet achieved.

Based on deep learning, it may never be possible for LLM studies to achieve AGI. This aligns with the view presented in "The Master Algorithm" by Pedro Domingos, Professor Emeritus at the University of Washington. He suggests that no single algorithm or AI model will lead to AGI. Instead, it could be a collection of interconnected algorithms that combine different AI approaches leading to AGI.

Nevertheless, Anthropic seems to have jumped to the forefront of LLMs for now. The company took a daring stance with bold statements about Claude's understanding capabilities. However, real-world adoption and independent benchmarking will be needed to confirm this position.

Despite the seemingly cutting-edge nature of today's news, it's important to consider the rapid pace of progress in the artificial intelligence industry. When the next stage arrives and what it will entail remains unknown.

In Davos in January, Sam Altman stated that OpenAI's next big model "will be able to do much, much more." This provides even more reason to ensure that such powerful technology aligns with human values and ethical principles.

Gary Grossman is the Tech Training VP at Edelman and a global leader of Edelman AI Center of Excellence.