I would like to thank all those who have shown an interest in my analysis of how ChatGPT will cause radical changes in the world as we know it, both publicly and privately. Of course, I wanted to be provocative, but above all, I wanted to provoke thought around how each of us, the institutions we respect, or even our work may be affected. I have experienced several "waves of Artificial Intelligence" that created small earthquakes in the translation industry, or even the language industry, despite these being small industries compared to others.
There are many people who have no need for machine translation and are not familiar with Google Translate or DeepL, for example. ChatGPT, however, is present in the press in almost every country in the world.
In this article I want to touch on some issues and concerns that I decided to hold back
on in the previous one and that I hope will help keep the debate alive.
Text generation and fake news
OpenAI has already expressed concern about the generation of fake news and misuse of
the tool: by increasing text generation capabilities, we are also increasing the capacity to generate fake news, which can be distributed with "human quality" and infiltrate results
on social networks. ChatGPT can write a story using Shakira's or Barack Obama's style
(in a recent internal test, our CTO generated an article on data anonymization that summarized the basics and opened as follows: "My fellow Americans, Today, I would like
to talk to you about an important issue that affects all of us: data masking").
For me, the biggest advantage is its ability to summarize: we cannot read 100 books on a specialized topic, but ChatGPT can generate the summary for us. The summary it generates is larger and more in-depth than your typical abstract. You can ask for a list of 10 books on business management and the recommended Top 3. With just two pages, you can judge whether the whole book is of interest to you and decide whether to purchase it or not. That is to say, it can act as an endorsement, and we all know that where there is endorsement, there is influence that can lead to purchases.
As is currently the case with the revision of machine translation by translators, ChatGPT can generate articles that you just need to correct and supplement with your "added value" in the form of updated information and world knowledge. With machine translation, only translators who translated like robots had a hard time. Good creative translators, capable of thinking, co-creating and providing cultural value, outperform the machine and are still making a lot of money.
If there is a margin for error, as in creative (pre)processes, we have a tool that increases our capabilities and is great.
Panic at Google
A leak indicated that Google's CEO called for emergency meetings to gather huge sums of money to "stop this right now." One of Google's advantages has always been solving problems as fast as possible. With Yahoo!, it took 40 seconds to answer a question and that was a real obsession for the company. The model is based on monetizing ads according to our search trends, which has always surprised me, as do the thousands of SEO agencies that know how to manipulate the algorithm so that certain websites appear on the first page and do not get buried on the second or third one. In my opinion, that is far from the true search for knowledge or recommending a book or a movie.
But with ChatGPT we are going to have an emerging conversation, we are going to stay and ask for more information. Maybe we will end up searching for more specific information or asking for help. Google rose to fame very quickly at the expense of Yahoo!, so it now faces the same danger. Sam Altman, the CEO of OpenAI, has stated several times that nothing exciting has happened in the world in the last 15 or 20 years except for "search." My bet is that there will be a hybridization between searches and information contribution from a language model (in fact, as I write these lines, I have already qualified as a tester of the Bing Chat system, whose technology seems to have been "controlled" to avoid certain delusions). The search engine you.com has been doing the same thing for some time now. Google became Google thanks to advertising. If it changes and becomes something else, perhaps including search trends in the Chat itself, there will be competition. That is why Microsoft shelled out $10 billion in the blink of an eye to integrate GPT-3 into Bing.
All this is a huge spanner in the search giant's works. Our relationship with Google is input-output, not cognitive. When we use a social network, we can have a good or bad experience, but it is not cognitive either. We know that we are using a tool and our brain is not affected. However, as with virtual reality, despite the fact that we "know" that what we are seeing is not real, it affects us (hence the "reality" part), because our brain perceives it as a cognitive experience. ChatGPT provides us with information in a cognitive way; it is forceful and therefore credible, even if it contains errors. We know that it is not human, but we perceive it with human realism and that provokes effects of attraction and companionship on our brain.
However, search and advertising are not going to be the object of the war between tech giants, rather it will be Artificial Intelligence itself. We are facing disruption that will change everything for good, including all the tech giants. We know that Apple has announced its own search engine and that it never settles for less than excellence and perfection. Maybe there will be some surprising news this year.
Winners and losers
Amazon has spent years on making Alexa a faster way to access information, giving it control of your life in the sense of home security, thermostat, all home appliances, etc. Beyond being a good alarm clock, it has not really worked out. People have given it a chance, but have ended up opting for their mobile phone. But what if ChatGPT is faster than a mobile phone in matters of voice services in our homes? Could we become surrounded by ChatGPT? Alexa's fate is similar to that of Siri, although the latter began with the advantage of having a whole closed system behind it.
Well, neither Alexa nor ChatGPT are AI. These are assistants, robots like Her with human voices. In fact, ChatGPT, using its security rails as it should, insists upon the fact that it is a "language model" and is not trained to give this or that answer.
Now, if with Microsoft's VALL-E service we can clone our voices in 3 seconds to work with other generative AI models, such as GPT-3, it means that a large language model like ChatGPT would soon be able to offer voice results, once this model is integrated. It is not a futuristic daydream. The question is how to bring it to millions of users in an ethical way, and OpenAI makes one thing clear on its website: they are concerned about the ethical uses of AI, exposure to minors, etc.
Which brings me to the next question: if there is concern around young people's mental health because of their overexposure to technology, what's going to happen now? Going back to the feelings developed in the movie Her, are we going to prefer interaction with machines over our fellow humans? The reality is that we are already talking to machines more and more every day, and perhaps we prefer it to some extent (for reasons too long to expose in this article). Soon, we may be asking ourselves these questions: how much more time will a young person spend talking to a language model like ChatGPT than talking to their mother or father? Is this bad? In reality, it is inevitable. Many of us remember childhood and many classic TV series and cartoons as experiences we didn't have with our parents, but as young spectators. The new generations communicate at a distance, maintain affective and family relationships at a distance, often using social tools. Again, new conversational technology changes not only the way we as humans relate to machines, but also the way we relate to each other. Intelligent interactions will shift from being human-to-human to incorporating human-machine cognitive interactions.
GPT4 and March 2024
People who have tried it say that it is a much bigger leap forward, even bigger than the one from GPT-2 to GPT-3 and 3.5 (ChatGPT), trained with 175 billion parameters.
Next year, we will see what version 4 will bring, with trillions of parameters and 500 times more powerful than ChatGPT. What is happening now is a mere experiment/toy. With integrations that we will see in IoT, possibly with voice, we are going to feel like we are interacting with a person, not like the artificiality of a call center. It could be a search engine/recommender with assistant functionalities and far more complete information than the small summaries it gives us now.
More paid versions may also emerge, because the computational cost of running a large language model is five or six times more than a search engine. Will some people have access to information and others not be able to afford it? Are we ready for what is coming?