How AI chatbots can revolutionize internet search
So far, Microsoft has only managed to attract a small number of users with its Bing search engine. But now, the artificial intelligence behind the text chatbot ChatGPT is expected to help this Google competitor achieve a breakthrough. But how does AI actually work? Here are the most important questions and answers.
Google has had an unassailable lead in Internet search until now. Now Microsoft presented a new search engine with artificial intelligence that is supposed to break Google's dominance. This could not only change the balance of power on the web, but also the way people search the internet.
Why are Google and Microsoft suddenly so active in AI?
In fact, Google and Microsoft, as well as other tech companies like IBM and Salesforce, have been working for years to improve their products with artificial intelligence features. The spectrum ranges from the automatic translation of texts to the calculation of travel routes in a navigation app.
However, the topic of AI has gained a whole new visibility since November 2022. At that time, the Californian start-up OpenAI made its AI language models with a text robot available to the public, even though they were not yet ready. OpenAI CEO Sam Altman's daring was rewarded. Use of ChatGPT exploded. The system is on everyone's lips.
What's so innovative about systems like ChatGPT?
The basic technology of ChatGPT was developed five years ago, primarily by Google researchers. It involves analyzing huge amounts of text using algorithms that mimic nerve cells in the brain. These "neural networks" can then recognize which words frequently occur together, for example "Hanover," "state capital" and "Lower Saxony. This enables ChatGPT to predict with a high degree of probability which word will occur next in a sentence and thus formulate plausible-sounding sentences.
What is new is that different AI functions can be combined with each other in the process. For example, ChatGPT can be asked to write a specific topic for an entry in a career network such as LinkedIn in English with a length of 300 words.
What specific applications are conceivable?
Systems like ChatGPT can aggregate information from a wide variety of sources. For example, when searching for a travel destination that is no more than four hours away by plane, one does not receive links to offers from travel portals. Instead, the chatbot delivers a text formulated by the AI in which Marrakesh, Lisbon, Faro, Istanbul, Crete and Reykjavík are suggested as possible destinations, each with a brief description.
How has the technology worked for search engines so far?
The leading search engine Google has so far two methods to answer queries. On the one hand, there is the "Knowledge Graph", which is a knowledge pool that Google can display directly using sources such as Wikipedia and many databases. The core business, however, is the display of Internet links that match a search term. The order is decided by Google's algorithm, which is kept secret. An important factor is the popularity of the site, i.e. how often the offer is linked. Parallel to these generic search results, Google displays links paid for by advertisers.
Could Bing become the new Google?
At least that's what Microsoft CEO Satya Nadella hopes. In recent years, Bing has never managed to seriously compete with Google. Google holds a market share of nearly 93 percent worldwide, according to Statcounter.com, while Bing comes in at just three percent. "This technology is going to reshape pretty much every software category we know," Nadella said. This also explains why his company has poured billions of dollars into OpenAI and the rights to use ChatGPT.
However, Google will not leave the field to its arch-rival Microsoft, with whom it competes primarily in the cloud business, without a fight. One day before the Microsoft event, Google CEO Sundar Pichai announced a comprehensive AI product initiative.
What are the issues with using AI chatbots?
The linguistic eloquence of the answers gives the impression that all the facts are also watertight. But this is not the case. ChatGPT repeatedly spits out answers that are fictitious. Even the makers of the chatbot warn against accepting the results unchecked.
Chat systems have particularly great difficulties when searching for people. It can happen that different people with the same name merge into a new virtual person. For example, the system might mistakenly believe that the CSU politician Gerd Müller was also the person who scored the winning goal for the German national team in the 1974 World Cup final.
Can AI bots also be abused?
Yes, this is possible. It is true that providers try to block malicious requests - for example, for instructions on how to build a bomb. But on the web you can see examples again and again of how these blocks can be circumvented.
ChatGPT and its counterpart "Bard" from Google are also suitable for cheating on exams and homework at school and university. So far, there is no effective way to clearly identify AI-generated content as robot text.
Are AI bots getting better and better?
Yes, they will. The opening for mass use of ChatGPT also contributes to the quality improvement. In fact, the creators at OpenAI ask users to rate the quality of the responses. This feedback helps the AI to identify systematic weaknesses and eliminate them. (Analysis: SDA)
