Is ChatGPT king? How high free AI chatbots fared throughout discipline testing


Whereas OpenAI’s ChatGPT was the primary synthetic intelligence (AI)-powered chatbot to captivate the world after its public launch in November 2022, quite a lot of opponents have entered {the marketplace} since then.

Tech giants Google and Microsoft have launched their AI chatbots, with Google’s Bard eradicating its waitlist, and opening up to over 180 countries and territories on Could 10, after Microsoft beat it to the punch and absolutely launched its AI-powered Bing search engine on Could 4.

With a number of chatbots to select from, Cointelegraph determined to place a few of the most well-known by their paces to see which held up greatest throughout discipline testing, in addition to evaluating a few of their options.

To check the chatbots, they have been every requested a collection of questions, riddles and extra complicated prompts to find out their accuracy and pace of responses.

Many AI chatbots out there in the present day are powered by OpenAI’s GPT fashions. Whereas these AI chatbots could give comparable outcomes to ChatGPT, the app builders also can add further instructions, which can change the outcomes.

OpenAI’s ChatGPT-3.5

Whereas OpenAI has already launched ChatGPT-4, which is out there to Plus plan customers for $20 per thirty days, ChatGPT-3.5 is free to make use of and is examined right here.

ChatGPT-4 significantly outperforms its predecessor with quicker response speeds, extra correct responses and fewer server downtime.

The primary AI chatbot to take the world by storm might help with duties like essay writing, code debugging and even private funds after solely a second or so of processing time.

Nevertheless, one space the place ChatGPT underperforms is its lack of potential to look the web.

This implies the mannequin is simply pretty much as good because the coaching information fed into it, which works up till September 2021. OpenAI is rolling out plugins that enable it to supply on-line info utilizing Bing’s search API, however this might be restricted to customers on the Plus plan.

Regardless of this shortcoming within the free model, the chatbot continues to be often in a position to recommend sources to assist the person with their question, as highlighted within the interplay beneath.

A screenshot illustrating ChatGPT-3.5’s incapability to talk of current occasions. Supply: OpenAI

ChatGPT-3.5 appropriately answered many of the riddles it was given and all the easy math issues, however the solutions have been much less constantly right when it was requested extra complicated issues.

For instance, when requested to resolve the quadratic equation 2t^2 + 0.3t – 0.4 = 0, ChatGPT-3.5 returned the right reply in a single out of three makes an attempt and had comparable points multiplying bigger numbers.

ChatGPT-3.5 may also be inaccurate when answering different questions. In keeping with OpenAI’s testing, it was solely in a position to appropriately reply 213 of 400 questions within the Uniform Bar Examination, which graduated legislation college students in america are required to move earlier than they will turn out to be working towards legal professionals.

Exterior of factual inaccuracies, ChatGPT-3.5 additionally struggled with questions to check its logical potential, such because the one beneath.

ChatGPT incorrectly solutions a query aimed to check its logical potential. Supply: OpenAI

Microsoft’s Bing

Bing’s ChatGPT relies on the GPT-4 language mannequin created by OpenAI, however the two chatbots have a number of key variations.

The primary noticeable distinction is that it takes Bing’s chatbot for much longer to reply to questions, with a mean response time of roughly 5 seconds in contrast with OpenAI’s ChatGPT taking just one second.

It additionally requires customers to make use of the Microsoft Edge net browser, which is nowhere close to as popular as Google Chrome.

On the constructive facet, Bing’s chatbot makes use of the Bing search engine in its responses, permitting it to reply questions on present occasions, not like some other chatbot utilizing GPT-4. It’s additionally at the moment out there at no cost.

Moreover, it supplies sources for its solutions, letting customers extra simply confirm claims made by the chatbot.

Microsoft’s Bing ChatGPT in motion. Supply: Bing

Utilizing the identical quadratic equation 2t^2 + 0.3t – 0.4 = 0, Bing linked to Microsoft Math Solver however usually gave an incorrect reply and had comparable points appropriately answering bigger multiplications.

In the identical logical query concerning the bookmark posed to ChatGPT-3.5, Bing appropriately answered that you’d anticipate to see the bookmark on web page 120.

Google’s Bard

Google’s not too long ago launched AI chatbot known as Bard, which runs on its PaLM 2 language mannequin.

As identified in a Twitter thread by AI fanatic Moritz Kremb, it could actually each reply and be prompted with pictures, helps quite a few programming languages and, like Bing’s chatbot, can connect with the web.

When requested how PaLM 2 compares with GPT-4, Bard stated that GPT-4 is best at producing textual content, however PaLM 2 is best at reasoning and logic, including:

“In the end, the most effective language mannequin for you is determined by your wants. When you want an LLM that’s robust at reasoning and logic, then Palm 2 is the higher selection. When you want an LLM that’s quick, good at producing textual content and has proved itself, then GPT-4 is the higher selection.”

Bard appropriately answered the bookmark query and it defined its reply in additional depth than Bing, however the explanations have been usually nonsensical.

Associated: What is Google’s Bard, and how does it work?

It solved many of the riddles it was given and carried out effectively on the maths questions, appropriately fixing the complicated multiplication questions and the quadratic equation in two of the three draft solutions it ready.


Whereas it additionally makes use of OpenAI’s GPT-3.5, there are some variations between’s YouChat and OpenAI’s ChatGPT.

It lists sources for many of the textual content it generates and in addition supplies hyperlinks to a number of net pages associated to the question.

It additionally connects to the web, permitting it to entry present occasions, and since it doesn’t have the identical degree of recognition as OpenAI’s chatbot, downtime shouldn’t be a problem.

It incorrectly answered each the bookmark query, the quadratic equation and the extra complicated multiplication downside.

It was in a position to resolve many of the riddles given to it however incorrectly answered some.


HuggingChat is an open-source AI chatbox from the AI agency Hugging Face, launched in April.

Requested to resolve the identical quadratic equation, HuggingChat returned 684 phrases of textual content and failed to supply a solution to the query. Whereas it may appropriately reply easy issues, it couldn’t multiply bigger numbers.

Whereas it typically gave direct solutions, HuggingChat usually returned huge partitions of textual content, which have been related initially however devolved into one thing akin to rambling.

For instance, it was requested to resolve the next riddle: “A barrel of water weighed 60 kilos. Somebody put one thing in it, and now it weighs 40 kilos. What did the particular person add?”

The proper reply is a gap, however the HuggingChat replied ice cubes earlier than launching right into a 545-word monologue.

What about the remaining?

There are various different AI chatbots at the moment out there, designed for extra restricted use circumstances than those talked about right here, with the market prone to proceed rising quickly.

For instance, Socratic is one other AI chatbot from Google that may be downloaded onto a smartphone to assist customers reply questions on science, math, literature and extra. It additionally supplies visible explanations of ideas in numerous topics and is a useful gizmo to help studying.

DeepAI is an AI chatbot that focuses on writing textual content similar to programming code, poems, tales or essays.


Whereas it could be unfair to check OpenAI’s ChatGPT-3.5 to Bing’s AI chatbot — given they’re utilizing completely different language fashions — this text intends to solely take a look at AI chatbots out there at no cost.

By means of Bing, customers can benefit from OpenAI’s ChatGPT-4 language mannequin, which is a large enchancment from its predecessor.

Whereas Google’s Bard was promising, Bing usually carried out the most effective of the present freely out there AI chatbots, however nonetheless made some errors.

Different chatbots seem to have extra restricted use circumstances that might be extra helpful, however these three appear to paved the way as improvement progresses.

Journal: Cryptocurrency trading addiction — What to look out for and how it is treated

The above represents an off-the-cuff discipline testing of various AI options and is certainly not exhaustive or consultant of Cointelegraph’s place on a selected AI answer.