While ChatGPT is in the news, Google is working on a similar chatbot. Indeed, while OpenAI’s artificial intelligence is cause for concern for many workers, it is also a cause for concern for Google, according to rumors. Although ChatGPT is still full of imperfections, its performance makes it a potential contender for the search engine. Indeed, if this AI continues to develop, its answers could gradually replace Google’s search results.

But the Mountain View firm does not intend to be overshadowed. This one would have already declared a “red code” and mobilized engineers to react to the launch of ChatGPT in 2022. And obviously, these rumors were founded since Google plans to launch a chatbot comparable to ChatGPT this year. This was revealed by Demis Hassabis, the CEO of DeepMind, in an interview with Time.

As a reminder, DeepMind is a subsidiary of Alphabet (Google’s parent company) specializing in artificial intelligence. DeepMind’s chatbot will be called Sparrow and it plans to launch the private beta this year. And since DeepMind and Google are part of the same entity (Alphabet), it is possible that Sparrow will become the competitor of ChatGPT by Google (especially if Microsoft integrates ChatGPT on its products and services).

DeepMind vs OpenAI ?

Nevertheless, if DeepMind will therefore show what it is capable of, the subsidiary of Alphabet wants to adopt a more cautious approach. According to Time, Sparrow’s delay is necessary for the company to work on features ChatGPT lacks, such as attribution. DeepMind will also try to establish limits that should not be crossed for its artificial intelligence. In a publication featuring Sparrow in September 2022, the Alphabet subsidiary described it as: “a dialogue agent that is helpful and reduces the risk of dangerous and inappropriate responses.”

“Our agent is designed to speak with a user, answer questions, and search the internet using Google when it’s useful to find evidence to inform their answers,” DeepMind also said in its post. This would have determined a set of rules to ensure that the “model behavior is safe”. These rules include a ban on making threatening statements, or making hateful comments. DeepMind also prohibits its AI from “pretending” to have a human identity.

DeepMind seems more cautious

In September, the subsidiary of Alphabet admitted that for the respect of these rules, Sparrow still has progress to make. “After training, participants were still able to trick him into breaking our rules 8% of the time, but compared to simpler approaches, Sparrow is better at following our rules in an adversarial poll”had indicated DeepMind.

As for the quality of answers, tests presented by DeepMind in September showed that 78% of the time, Sparrow was able to answer factual questions plausibly and with supporting evidence. When the chatbot is in private beta, we can start comparing it with ChatGPT, in terms of the quality of the answers, but also with regard to ethics.
Note that for its part, OpenAI has already warned that although ChatGPT can impress with its answers, the artificial intelligence still needs improvements.

Source link