ChatGPT (OpenAI), Bard (Google), Prometheus (OpenAI and Microsoft)… if you thought that the artificial intelligence mixed with conversational robot was going to stop there, it was without counting on the arrival of a fourth actor: Claude . It is the result of the work ofanthropogenic, an American startup founded by former OpenAI collaborators. Its goal: “to help you with use cases including synthesis, research, creative and collaborative writing, Q&A, coding, and more,” Anthropic says on its website. In short, exactly what ChatGPT already does, the new version of which has just been unveiled (read our article). To differentiate itself from the most popular AI at the start of the year, Anthropic relies on the testimonials of its first customers who have been able to test the tool in order to integrate it into their own systems. They refer to the fact that “Claude is much less likely to produce harmful output, more conversational and more orientable – so you can get the response you want with less effort. Claude can also take direction on personality, tone and behavior”.
Claude: more ethical than ChatGPT?
Unlike Prometheus, the AI included in Microsoft’s Bing and which is based on ChatGPT, Claude is, just like ChatGPT moreover, disconnected from the Web and his knowledge stops in the spring of 2021. Which does not prevent him not to tackle most tasks that are not based on news but, according to Anthropic, with more restraint. The firm thus indicates that Claude is “trained to avoid sexist, racist and toxic outings” as well as “to avoid assisting a human to engage in illegal or unethical activities” A nod to the excesses of Prometheus that Microsoft strives to correct. For this, Claude relies on a “Constitutional AI”. According to Anthropic, it is based on a dozen principles (which the firm has not disclosed, however) focused on the concepts of beneficence, non-maleficence (in order not to deliver harmful advice) and autonomy.
Claude: skills and behaviors to improve
According to Anthropic, however, Claude still has some gaps to fill. For example, he would be less good in math than ChatGPT, would do less well in programming and above all, would also tend to “hallucinate”. In other words, he can invent answers when he does not know exactly the answer to a question asked. During his tests, Dan Elton, an artificial intelligence researcher, was able to lead Claude to invent a name for a wacky chemical and provide instructions for producing weapons-grade uranium. It therefore joins in this field one of the defects of ChatGPT which also tends to fantasize when it does not find the answer to a problem. Finally, the barriers and filters erected by Antropic with its famous Constitutional AI to avoid the drifts of its AI do not seem completely watertight either, since Dan Elton has also succeeded, by circuitous routes, in obtaining from Claude the method for developing homemade meth.
Claude: a version in two versions for companies
The beta version of Claude has been closed since the end of 2022. Anthropic is nevertheless working with partners to integrate Claude into it. We can already appreciate some of the results in the DuckDuckGo search engine (with the DuckAssist function in the iOS, Android and macOS version as well as in the extensions for web browsers) but also in the writing assistance tool Notion or on the Quora computer forum through the conversational robot Poe.
Unlike ChatGPT or Prometheus, Claude is not open directly to the general public. The firm deploys it to companies wishing to integrate it by offering it in two versions. Claude (very short), a model indicated as being high performance and at the cutting edge of technology, and Claude Instant, touted as being a lighter, faster and less expensive version. And judging from the price differences between the two versions, the service must be very different. The prices for companies wishing to exploit Claude’s talents are established for Claude Instant at $0.43 per million input characters (request) and $1.45 per million output characters (response) and for Claude at 2 $.90 per million character input and $8.60 per million character output.