Google has announced that it will launch a competitor to OpenAI’s ChatGPT model, which has been very popular lately. Google’s new chatbot is called Sparrow and was cited as supporting evidence in a research paper last year. DeepMind CEO Demis Hassabis said that Sparrow could launch in private beta in 2023.
Sparrow’s project has been delayed since the launch of ChatGPT due to the risks associated with this type of generative technology. These risks include the ability to generate malicious and biased texts, as well as scammers using the model to create phishing emails and malicious code.
Sparrow wants to be more reliable
Another important consideration for DeepMind is the accuracy of the information. Language models like ChatGPT generate responses that may appear legitimate but have not been verified for accuracy. To remedy this, Sparrow should have important features like the ability to cite certain news sources.
According to the research report on how Sparrow works, he backed up his plausible answers with evidence 78% of the time. DeepMind is also working on rules to constrain Sparrow’s behavior and ability to rely on human input in certain contexts.
It should be noted that the above innovations are important steps for this type of tools. By adding links to cited sources, Sparrow tastes more like Google than ChatGPT. It remains to be seen how this will address concerns in areas like education. However, citing the sources used to formulate the answers could provide an excellent starting point for research, but could also look like a finished essay with appropriate references.
Google’s business model at risk
Google is entering dangerous territory by launching a competitor to OpenAI’s ChatGPT, since its entire business model relies on advertising revenue from websites. If search is mainly done by ChatBots in the future, their entire business model could collapse. While Google could add advertising to Sparrow, that would be fewer than the current model, which heralds its long-term demise.
Additionally, Google could fall into the trap of the inventor’s dilemma, where, despite its superiority in AI, it could be outperformed by smaller but much more flexible companies like OpenAI. It is important for Google to find new business models to maintain its position in the AI industry.
ChatGPT powered by Wolfram Alpha
Regarding ChatGPT’s accuracy, it is true that it can run into difficulties on certain domains. He’s like a kid pretending to know everything. He there are ideas to connect ChatGPT to Wolfram Alpha, a pure data and computational search engine.
Wolfram Alpha is like a genius scientist who can communicate with the ChatGPT kid in such a way to provide users with relevant and more accurate answers. This could greatly improve the accuracy of the responses generated by ChatGPT. It is important to note that this idea of connecting ChatGPT to Wolfram Alpha is still in development and not yet available to users.