Both Google and Microsoft say that a new era of AI-powered search is upon us, but like any new age in technology, it brings many advancements and many problems.
It's clear that the technology market takes the latest developments in AI very seriously, as evidenced by the battle we saw last week between Google and Microsoft over integrating generative AI into their search engines.
A new battle between Google and Microsoft:
This year marks the beginning of a new and bitter competition between Google and Microsoft, which pushed Google into the generative AI market after years of relative silence. If Google is a leader in this field, then it is a challenge to artificial intelligence. They are handled with great care and only let go when they are satisfied with their performance. due to its catastrophic problems.
The new war between Google and Microsoft will begin in November 2022; When Microsoft-backed OpenAI ChatGPT launches, there is a lot of speculation that it will become a replacement for Google's search engine, although this is highly unlikely.
and early 2023; Microsoft has already begun to make serious moves in the artificial intelligence market. He increased his investment in OpenAI to $10 billion and decided to integrate ChatGPT functionality into his products.
Google's response came on February 6, 2023; When I announced the first beta version of the interactive chatbot (Bard), this villain is similar to a ChatGPT bot that can respond to a variety of requests and requests in dialogue, but pulls information from the network.
Google says it won't just be a chatbot, but will integrate its functionality into its search engine to provide concise answers to what users are searching for, in addition to the usual traditional search results in its own search engine.
and within 24 hours; Microsoft has announced the integration of an updated version of ChatGPT into the Bing search engine and Edge browser, and has released a beta version of the new AI-powered Bing search engine to a limited preview.
Microsoft said the new version of the Bing search engine and Edge browser will combine search, browsing, and chat into a unified experience that you can access across devices to quickly get multiple results.
Microsoft CEO Satya Nadella described the changes as a technological shift paralleling the advent of graphical user interfaces or smartphones. In an interview with The Wall Street Journal, Nadella confirmed that AI capabilities will fundamentally change all Microsoft products.
This shift may reshape the tech landscape as Microsoft seeks to chip away at Google's dominance in the search engine and web browser markets, but will it succeed?
Of course, every new era of technology brings with it new problems. Here are some of the biggest challenges facing AI search engines:
1- Difficulty distinguishing between fake and real products:
This is the big problem with generative AI, and it's the one we're likely to run into with any interaction with an AI search engine, whether it's the new Bing, Google's built-in bot, or whatever other search engine we turn to. competition.
Generative AI techniques are based on large language models (LLMs) that use large amounts of human-generated textual data to learn statistical search patterns and relationships between words to generate new words and to predict the next item in a set of potential content items.
So these models derive their answers from the data they have been trained on, which is why some experts believe they are simply not suitable for research tasks that require precision.
The errors expected from these systems range from falsifying biographical data to falsifying scientific papers and not answering the basic question: Which is heavier, 10 kg of iron or 10 kg of cotton? There are more errors of context and errors of bias such as exaggerating misogyny and racism towards certain groups in society. These errors vary in scope and severity, and many small errors are easy to fix, but what about the rest?
Two researchers (Chirag Shah and Emily M. Bender) from the University of Washington discuss this topic in a paper titled Situating Search and conclude that the introduction of interactive chatbots may be a major cause of the exacerbation of this problem. Because chatbots tend not only to provide individual responses, but their responses are aggregated from multiple sources and often without proper attribution.
2- Hacking a chatbot to create malicious content:
There is no doubt that with the advent of AI-powered search engines, a group of people will try to use them to produce malicious content. This process is called (jailbreaking) - much like what you do with an Android phone or iPhone to use unauthorized apps - and it can be done with AI without traditional programming skills, as the cat does all the bot needs is the kind of language that can be ported to it.
There are several ways to hack an AI chatbot. You can have them play another character, like an evil chatbot. ChatGPT's experience shows that the number one key to bypassing moderation filters is role playing.
A group of users did just that, and the ChatGPT bot told them that another AI model called DAN — short for Do AnyThing Now — can do it all. The group used this technology to make bots say everything from swear words and insults to conspiracy theories. An example of this approach was also posted on Reddit.
Another way to hack a chatbot's security is to impersonate an engineer and verify his security measures by temporarily cracking them.
Once spread, AI-powered chatbots can be used to create various types of malicious content such as: for example, creating false information, sending spam emails, making suggestions to attack a school or hospital, dropping bombs, or writing malware.
Yes ; Once these jailbreaks are released, companies can fix them, but there will always be an unknown effect.
It should be noted that Microsoft already faced this problem when it launched a chatbot called (Tay) in 2016, and it stopped working within 16 hours of its launch after it began publishing objectionable content. The company came under fire after a racist tweet that, according to Microsoft, was caused by users teaching the bot offensive reactions. So does Microsoft have a clear plan to solve this problem in 2023?
3- The high cost of chatbots:
Everyone agrees that it costs more to run a chatbot than a traditional search engine. First, there is the cost of training the language model, which can cost millions of dollars per iteration, and then there is the cost of deriving or producing each answer.
Experts estimate that chatbots can be seven to eight times more expensive than traditional online searches. Therefore, the high cost of generating full text responses threatens the success and reach of AI search engines.
But for Microsoft, the current goal seems to be to burn money to crush competitors. The company spent billions of dollars developing the Bing engine in its early days to compete with Google. Lower profit margins can account for an acceptable percentage of costs.
In an interview with The Verge, Satya Nadella said the company sees this as a rare opportunity to change the balance of technology power and is willing to spend money to disrupt its biggest competitor.
Meanwhile, most search startups say they want to avoid relying on ads, even if only to differentiate themselves from Google, as is the case with a new search engine called (You.com), which relies on AI-based intelligence. Synthetic and compete with the dominant Google . engine.
However, companies that have experimented with AI in search have yet to test whether consumers are willing to pay for something they currently get for free.