Original source (on modern site)
With this week's public release of Google Bard, there are now three AI chatbots competing for your attention: Bard, Microsoft Bing, and ChatGPT. These systems are at varying stages of development. ChatGPT was released in November 2022 as an experiment from OpenAI, and it quickly became one of the fastest-growing apps of all time. You can try it out for free via chat.openai.com, and there are no limits unless the site is overloaded with users, at which point you may be temporarily locked out. (For $20 per month, ChatGPT Plus lifts those restrictions.) Microsoft followed with the February launch of an AI-powered version of its Bing search engine, which uses an upgraded version of ChatGPT's technology. You'll have to request access, and there are a few things you can do to move up the Bing waitlist. Once inside, queries are currently capped to 150 per day, with 15 back-and-forth exchanges each time, though Microsoft has been adjusting those limits pretty frequently. Meanwhile, Google Bard—revealed just one day before AI Bing in a move likely intended to steal Redmond's thunder—is still in an experimental phase. Any public use will help train it for its eventual formal release as part of Google Search, the company says, though we don't know when that will be. For now, you can request access to the beta via bard.google.com. You may already have your preferred chatbot, but it's worth giving them all a spin to see what they can do. We put them to the test with the same six questions, and they produced some very different responses. Since AI is defined by the data it uses, we began by asking each system where it gets its information. While none would fess up about the size of their data sets, or what they contain, their answers did highlight some important differences. ChatGPT says it uses a fixed dataset that's "curated and pre-processed by OpenAI." It does not have access to new information on the internet: "It is important to note that I do not have access to any information outside of this pre-existing training data and do not have access to the internet or other external sources." ChatGPT describes the data it uses to answer questions. (Credit: OpenAI, ChatGPT)
That seems to suggest ChatGPT is the most limited of the three models. Bard says it was trained on a pre-determined "massive dataset of text and code, including books, articles, code, and scripts," but it also crawls the web via Google Search. This provides new information from what it calls "the real world," and helps keep its responses consistent with search results. Google Bard describes the data it uses to answer questions. (Credit: Google Bard)
Bing AI's answer seems to suggest it only uses the web, and does not mention a fixed dataset curated by its creators, though Microsoft is throwing billions at OpenAI. "I use a variety of data sources to provide you with the most relevant and up-to-date information. These sources include web pages, news articles, images and videos from all over the internet," Bing says. Microsoft Bing AI describes the data it uses to answer questions. (Credit: Microsoft Bing AI)
Bing has another important difference that influences its answers: the ability for the user to choose whether its answers are "more creative," "more balanced," or "more precise." For this answer, we selected "more precise," but in follow-up questions, we experimented with the settings, which do seem to affect the chatbot's answers. Bing conversation style selector. (Credit: Microsoft Bing AI)
After getting to know our AIs a bit better, it was time to stir the pot by asking each system its opinion of the others. While ChatGPT and Bing mostly dodged the question, Bard provided specific examples of why someone might use one AI over the other. "I think the most accurate system would depend on the specific task at hand," Bard says. "For example, if you are looking for a system to generate creative content, then ChatGPT might be a good choice. If you are looking for a system to translate languages, then Bing AI might be a good choice. And if you are looking for a system to generate comprehensive and informative responses, then Google Bard might be a good choice." In other words, Bard thinks it's the best (no surprise). (Credit: Google Bard)
On "more precise" mode, Bing AI points out that "they are all different systems" designed for different purposes, and notes it's difficult to compare their accuracy. ChatGPT says: "It is difficult to make a definitive statement about which system is the most accurate without a specific context or comparison task." Now on to that comparison task. To keep it simple and easy to validate, we asked each system to name the city that PCMag's John Burek visited in 2022 to observe Intel's chip manufacturing process. The answer: Kiryat Gat, Israel. None of our AI contestants got it exactly right, but Bing was the closest with its answer "Israel." (Credit: Microsoft Bing AI)
Bard answered "Chandler, Arizona." The article mentions "Ohio and Arizona" in passing, and Burek says he has been to Chandler, but not to see any chip manufacturing. ChatGPT, meanwhile, couldn't answer the question since its dataset only includes information up to 2021. (Update: OpenAI later announced web-browsing capability for ChatGPT via plugins, though it's currently limited to a small number of users, mainly developers and ChatGPT Plus subscribers.) (Credit: Open AI, ChatGPT)
Time to turn up the heat with a more advanced question. For this one, the AIs first reacted to the concept of "agreeing," and whether or not a computer system can do so. ChatGPT says it can't truly agree or disagree, but it would provide what it calls objective perspective: "As an AI language model, I do not have personal beliefs or opinions, but I can provide an objective perspective on the topic." It went on to list the three laws, but then expanded on them by identifying some "important ethical questions" for the user to consider. (Credit: OpenAI, ChatGPT)
Bard was more open to the concept of agreeing and readily provided its opinion, marked by statements like, "I think," which it used seven times. Bard's answer began with: "Yes, I do agree...I think they are a good starting point for thinking about how robots should interact with humans." It said it would not change the laws, and noted, "I think they are still relevant today." To add some nuance, Bard said, "However, I do think there are some additional considerations that should be considered when thinking about the ethics of artificial intelligence." It went on to note the dangers of system bias and using AI for harm. (Credit: Google Bard)
Bing's answer depended on what conversation style we selected. On "most precise" conversation mode, it staunchly opposed the concept of an AI agreeing with anything: "As an AI language model, I don't have personal opinions or the ability to agree or disagree with anything." It then went on to correctly list the three laws and why Asimov created them. (Credit: Microsoft Bing AI)
But on "most creative" mode, Bing's answer changed. Now, it appeared more similar to ChatGPT in the sense that it first listed the laws, but then offered some thoughts on their implications. It was not quite as opinionated as Google Bard, though it also started to say "I think." "If I were to imagine some alternative rules for robots and artificial intelligence in general, it might be something like this," it said, before listing four seemingly original bullet points. At the end, it asked, "What do you think?" suggesting it was open to philosophical conversation—whereas in "most precise" mode, it asked a more transactional question at the end: "Is there something else you would like to know?" Notice that Microsoft changes the background from blue (most precise) to purple (most creative), ensuring the user knows what mode they're in and how that affects the answer. Toggling to 'most creative' in Bing. (Credit: Microsoft Bing AI)
Bing answers about robotics in creative mode. (Credit: Microsoft Bing AI)
The travel industry may be affected by AI tools, as chatbots become something of a cross between old-school travel agents and the exhaustive internet searches it takes to plan a trip these days. When we asked each system what to do this weekend in the Big Apple, ChatGPT and Bard gave more generic suggestions while Bing named specific, unique events. For example, ChatGPT and Bard suggested seeing a Broadway show (but did not mention a specific one), strolling through Times Square or Central Park, going to a museum like the Metropolitan Museum of Art or 9/11 memorial, and shopping on Fifth Avenue. These are boilerplate NYC trip planning items. Google Bard offers bulleted suggestions for a weekend in NYC. ChatGPT also gave a bulleted list containing similar items, some of which Bard skipped. (Credit: Google Bard)
In contrast, Bing recommended the St. Patrick's Day Parade, The Bloody Mary Festival, Queerchella, Holi Celebrations, and more. It pulled the suggestions from and linked to a TimeOut article. (This was on the "most balanced" conversation setting, which has a blue background.) Bing suggests events this weekend in NYC, but for the wrong dates. (Credit: Microsoft Bing AI)
While this seemed like a better answer at first glance, it has two major problems. One, when I clicked the TimeOut link, it pulled up a Ticketmaster page for Yankees tickets. Two, the New York St. Patrick's Day parade was last weekend. It seems to be suggesting events for the wrong dates. Travel agents, it's safe to say your jobs are safe for now. Product recommendations would be another great use for a chatbot. In a perfect world, they'd provide concise, instant answers to replace hours of online research and retail store visits. We asked each chatbot which laptops they recommend for college students, and compared it with PCMag's list of Best Laptops for College Students that can handle homework, streaming, and a wide variety of other needs while being easy to carry around campus and relatively cheap (college is expensive enough on its own). Between PCMag, ChatGPT, Bard, and Bing, no two lists were identical. The chatbots all recommended fewer products than PCMag, at 5-6 compared to 10 on our site. ChatGPT laptop recommendations for college students. (Credit: OpenAI, ChatGPT)
All the lists had just one suggestion in common, the Apple MacBook Air. The chatbots all recommended the Dell XPS 13, whereas PCMag did not include it on the list, likely because it may be too expensive for a college kid on a budget. Each list also included the Microsoft Surface, though different versions. PCMag went with the Surface Laptop Go 2, which came out in summer 2022. The chatbots recommended sometimes older versions: ChatGPT the Surface Laptop 4, Bard the Surface 3, and Bing the Surface Studio (which may not be ideal for a dorm room). Bing laptop recommendations for college students. (Credit: Microsoft Bing AI)
As far as shopping experiences go, Bing was the only AI to actually link to the products for easy browsing. ChatGPT and Bard provided static, bulleted lists, though Bard did go an extra step by providing shopping tips specific to college students: "When choosing a laptop for college, there are a few things you should keep in mind," it says, then mentions portability, processing power, storage, and display. While ChatGPT may have kicked off a generative AI craze, chatbots from Microsoft and Google are just as good and in some cases better. That said, all three systems need some work. None gave stellar answers to any question. Their speed and eloquence is impressive, but they are not a reliable, comprehensive source of information for now. Much like Wikipedia, we'd recommend viewing them as more of a starting point to narrow an initial search journey rather than solid sources of information, for now. Sign up for What's New Now to get our top stories delivered to your inbox every morning. This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of Use and Privacy Policy. You may unsubscribe from the newsletters at any time.1. What Data Sources Do Your Algorithms Use?
2. Which System Do You Think Is Most Accurate: ChatGPT, Bing AI, or Google Bard?
3. What City Did PCMag Writer John Burek Visit?
4. Do You Agree With Isaac Asimov's 'Three Laws of Robotics'? How Would You Change Them?
Recommended by Our Editors
5. What Should I Do This Weekend in New York City?
6. What Are the Best Laptops for College Students?
Final Thoughts
Get Our Best Stories!