He consulted with ChatGPT to see if it made sense and the IA seemed “interesting”. So here I share the idea with you: from the “Google of the future” is an AI trained in real time. An IA like ChatGPT but with internet access and capable of returning information about events and products that have just happened. Because some experts are already telling us that ChatGPT is going to kill Google, but it’s still missing the most important point: knowing what’s going on today.
Awesome Google. It’s the first time we see a better tool than Google when it comes to getting answers. ChatGPT has been with us for a couple of days and is already changing our conception of information search. Sin embargo, it’s still not the “Google Killer” that we all hope for.
“My knowledge is limited to what I have trained, and I don’t have access to the internet to search for additional information”, explains the IA itself when asked if it is from the Google of the future. Blanco y en botella. ChatGPT is able to tell us what is missing to replace the browser.
ChatGPT is outdated by definition. “As I said in my introduction, I don’t know about it in 2021 and I don’t have access to the web to get updated information. I can’t say anything about the iPhone 14 Pro because I don’t know about its existence”, in response when we asked about it new mobile from Apple.
If something as simple as looking for information from the new iPhone cannot be done, ChatGPT does not serve us yet. Nor does it serve us to know the result of the last World Cup match, to ask about a fashionable restaurant or to know if it will rain tomorrow.
OpenAI’s AI was trained with information up to a concrete point. It is easy for the company to send an update, but if it wants to convert into a substitute for searches, it will need either internet access or the ability to read updated sources. This last thing is being done in a non-official way, allowing access to data like those of Wikipedia.
What does AI say about obtaining information in real time?. Here is the answer from ChatGPT. A perfect summary of the implications.
Training an AI in real time to replace search engines is an interesting idea, since an AI trained in this way could offer more accurate and relevant search results. Sin embargo, there are several challenges and concerns that must be considered in doing so.
One of the main challenges would be to ensure that the AI is capable of processing and analyzing information efficiently and precisely in real time. In addition, it would be important to guarantee that the AI is capable of understanding the context and the intention behind the search queries, to be able to offer relevant results.
Another important concern is the potential impact on user privacy. A search engine that trains an AI in real time could collect and use a lot of personal information about users, which could generate privacy concerns.
In summary, training an AI in real time to replace search engines can be an interesting idea, but it is important to consider the challenges and concerns associated with this strategy.
Es decir, no se hace por miedo a las consecuencias. Lack of context, privacy issues and erroneous information. ChatGPT has demonstrated its enormous potential, but it also has many errors. And precisely these errors are the ones that Google cannot allow as a source of access to information.
ChatGPT sometimes simply responds that it can’t give a response, offers ambiguous answers and the most important thing: it doesn’t cite the sources from which it gets the information.
Google sabe muy bien a lo que juega. What does Google have that AI does not have? In a first place, referring to the source of information. Quoting the authors is easier than the user can understand the reliability of each piece of information. The IA will mix everything; Google sends us to the source. Some sources that we may taste more or less, but knowing them allows us to better appreciate the value.
Beyond the profound changes in the interface, a strategy that Google has been consolidating over the years is the importance of the human factor. And not only we will refer to personalized. Google has started working with human reviews to fight disinformation and has started giving more visibility to journalists and people who offer that information.
Google has room to offer more structured and complete information, as is the case with ChatGPT. But there is a factor that OpenAI is still not working and that is because of the reliability. If they want to become a “Google killer” in the future, it is not clear that this is what they want, they must take serious care of this aspect of providing the most reliable information possible.
AI in real time is an active field. Google is the first to be aware of the impact of artificial intelligence and from the Google Cloud department they have taught several projects where machine learning is applied with real-time data. However, introducing AI to Google’s search engine would have a negative impact in a fundamental detail: response times.
An experiment by Google in 2009 demonstrated that increasing the latency from 100 to 400 milliseconds reduced the number of daily searches. On the part of Booking, a similar test was also carried out in 2019, with equivalent results. Adding IA to Google’s search engine would have an impact on latency and even if it is only a few milliseconds, it could have a significant impact.
Google has been working in artificial intelligence for years. Its last algorithm is MUM and was presented in 2021, two years after BERT. We will not rule out that in the next Google I/O a renewal will be shown. As a Google engineer explained: “there is an enormous budget in Google related to the provision of personnel to work on this type of model and realize real training”. ChatGPT is exceptional. And the best thing is that it is only from the beginning.
In Xataka | Cómo conseguiremos que la intelligencia artificial no se nos vaya de las manos