Generative AI technology like Bard and OpenAI’s ChatGPT are still limited by the fact that they rely on human intuition to create believable narratives. However, that doesn’t mean they don’t have the facts wrong. They just don’t have the market share to matter.
The Trust Project is a worldwide group of news organizations working to establish transparency standards.
Although generative AIs like Google’s Bard or OpenAI’s ChatGPT have impressive capabilities in creating natural-sounding writing, they have already demonstrated the present constraints of this technology and their limited knowledge of factual information. For instance, Bard wrongly stated that the JWST was the first telescope to capture a picture of an exoplanet; however, this feat was accomplished by European Southern Observatory’s Very Large Telescope in 2004.
And yet, despite these shortcomings, the market for generative AI products is vast and promising, and companies are eager to get their products into the hands of consumers as soon as possible, willing to overlook a few errors in the process.
Anthropic’s team, consisting mostly of former OpenAI employees, has a different strategy for their chatbot, Claude. As per a report from TechCrunch, the team’s pragmatic approach has resulted in an AI that is more controllable and less prone to producing negative results compared to ChatGPT. This is exciting news for the industry.
The closed beta development of Claude has been ongoing since late 2022. The company has now started testing its conversational abilities with launch partners like Robin AI, Quora, and DuckDuckGo. Though the pricing details are not out yet, the company has confirmed that at the launch, two versions of the product will be available: the standard API and a lighter, faster version called Claude Instant. It’s an exciting time for the company and its stakeholders.
Richard Robinson, the CEO of Robin, shared with TechCrunch that the company relies on Claude for evaluating specific sections of their contracts and proposing alternative, customer-friendly language. Robinson expressed his enthusiasm for Claude’s exceptional grasp of language, even in technical areas like legal jargon. Additionally, Claude’s drafting, summarizing, translation, and simplification skills have greatly impressed the team at Robin.
Anthropic is confident that Claude will not exhibit the same behavior as Tay, who started spewing racist language. This is mainly due to their AI’s unique training program known as “constitutional AI.” The company claims this approach is based on principles that foster ethical alignment between humans and robots. While Anthropic has not revealed the 10 foundational principles, they have assured that these principles are centered around the concepts of beneficence, nonmaleficence, and autonomy. This marketing strategy of keeping the principles a secret seems to be working in favor of the company’s buzz.
The organization decided to create a distinct artificial intelligence system that could produce text in accordance with its semi-secret principles. This system was developed by responding to various writing prompts that included tasks like composing a poem in the style of John Keats. Claude was then trained using this model. Even though Claude is designed to be less problematic than its competitors, it still has the tendency to fabricate facts, just like a startup CEO on an ayahuasca retreat. This AI has even created a new chemical and taken creative liberties with the uranium enrichment process. Reports suggest that Claude has a lower score on standardized tests for both math and grammar when compared to ChatGPT.
According to the spokesperson from Anthropic, the greatest obstacle is to create models that avoid hallucinations while remaining practical. There is a risk that the models might choose not to say anything at all to avoid lying, which is a tradeoff the team is currently addressing. While the company has made headway in reducing hallucinations, there is still more work to be done. The spokesperson sounded thrilled about the progress made so far.
Read more related articles:
- ArtStation users are outraged that AI-generated content has not been removed
- Claude: The first ChatGPT competitor has arrived
- Best 10 AI Prompt Guides and Tutorials for Text-to-Image Models: Midjourney, Stable Diffusion, Dall-E
Any data, text, or other content on this page is provided as general market information and not as investment advice. Past performance is not necessarily an indicator of future results.