Published January 18, 2023
Author: Ash Khan

DeepMind intends to release a chatbot with moral constraints

DeepMind, a Google company, claims it plans to introduce a ChatGPT competitor shortly. They state that their chatbot promises to be a safer type of AI helper.

Introducing Sparrow

Sparrow was revealed to the public last year as a proof-of-concept in a research paper. The paper characterized it as a conversation agent that is beneficial and decreases the chance of hazardous and improper replies.

There are some reservations about the potential perils of chatbots, which DeepMind describes as inaccurate or created information. On the other hand, Sparrow appears to be ready to take flight in beta form shortly. Given DeepMind’s tight ties with workspace google’s parent company, it could become the search engine’s response to ChatGPT.

DeepMind and Google

DeepMind was bought by Google Workspace‘s parent company nine years ago. It has been a pioneer in AI research for the previous decade. The current focus is on Microsoft Office 365 parent company’s prized product ChatGPT. However, DeepMind is considering launching its chatbot, Sparrow, for a “private beta” in 2023.

DeepMind attributes the short delay in Sparrow’s release to ensuring that it contains critical functionality that ChatGPT lacks. Most notably, quoting particular sources. Hassabis told Time that “It’s right to be cautious on that front”.

According to DeepMind’s research report, Sparrow will be more limited and cautious at first than ChatGPT. The latter has gone widespread due to its outstanding ability to assist everyone from hackers to armchair poets. However, there are also some raised concerns due to its discriminating comments and malware-writing abilities.

DeepMind has talked about the behavior-constraining principles that Sparrow is based on. Additionally, it is willing to defer to humans in “contexts where humans can make better decisions. In early tests, Sparrow appears to deliver convincing answers. Significantly, supported it with proof of 78% success when posed a factual question.

However, its real capabilities will be revealed when the public beta is released later this year. We have to wait for the AI chatbot discussion between the Google-affiliated Sparrow and the Microsoft 365 office parent company ChatGPT. 

AI chatbots are still in kindergarten

Anyone who has used ChatGPT would know that it is capable of emulating intelligence on a variety of issues. While that is certainly entertaining, AI chatbots also require moral intelligence and the capacity to cite sources. DeepMind claims its Sparrow is a ‘conversation agent’ and excels in these areas.

Sparrow’s public beta is coming soon in order to take it to the next level. DeepMind argues that building better rules for an AI assistant “will involve expert input on many issues. These experts include legislators, social scientists, ethicists, and participatory input from a large array of users and impacted groups.

Similarly, the CEO of OpenAI which invented ChatGPT has spoken about the challenges of opening AI chatbots without creating collateral harm. The CEO claims there are challenges with the adoption of AI tech over time; we are not able to predict every difficulty.

In other words, ChatGPT and DeepMind’s Sparrow engineers are like parents with inquisitive kids. Both offer joy and peril – especially when their kindergarten instructor is the whole internet.

ChatGPT is on its way to becoming monetized with the impending release of Professional, a paid-for tier. However, DeepMind’s Sparrow is a more mild-mannered persona that AI chatbots require as they rush toward next-generation models like ChatGPT-4.