Artificial intelligence (AI) is one of the most competitive technological fields with the greatest projection. In recent months, technology companies such as Open AI, Microsoft, Google and Meta unveiled different products in a kind of battle for leading what could well be called “the creative revolution of machines”.
Big tech devised systems capable of generating original content from the billions of pieces of data with which they were trained. Platforms known as creative AI that generate amazement and concern for the impact they will bring to everyday life.
ChatGPT, the star of the moment
At the head of this battle stands OpenAI with its popular ChatGPTa language model that has up to 175 billion parameters, allowing it to generate, in most cases, accurate and consistent content.
“It was trained on general-purpose texts, which makes it useful for a number of different tasks, such as explain concepts, apply style, translate, paraphrase, etc, without the need to be specifically trained for each case. You are not connected to the internet so you do not have access to current events. However, OpenAI regularly adjusts it to incorporate that information”, stresses Agustín Di Salvo, leader of the artificial intelligence practice at Rocking Data, in dialogue with iProfessional.
To test the potential of this tool, just go to the official OpenAI page and log in. The chat can be asked to put together summaries or texts with material that is provided or to write texts based on the data that it is instructed to search in a timely manner.
Microsoft, seeing the potential in this project, invested billions of dollars in OpenAI and in February it announced the integration of ChatGPT to its Bing search engine. For now, only limited access to the system is possible. But that limited test was enough so that in a few days the networks were flooded with experiences that showed that, sometimes, the system could get out of control.
GPT Chat allows you to create unique texts on any subject in seconds.
Some users showed that they received aggressive answers, wrong answers and even The case of journalist Kevin Roose from the New York Times who said that the bot wanted to seduce him went viral and convince him to leave his wife. Apart from these anecdotal cases, the truth is that the system continues to evolve and is showing a lot of potential.
Bard, Google’s bet
On February 6 Google introduced Bard, an experimental conversational AI service that combines information from the web with the creativity and intelligence of language models. Bard is powered by the LaMDA language model that the company introduced more than two years ago.
He had his small setback when in the presentation prepared by the computer giant, Bard misresponded to a query. The mistake caused Google’s shares to fall about 9%, which is equivalent to a loss of value of the company in the stock market of 100 billion dollars. However, this does not imply that it is the end.
It has not yet been announced when this tool will be made available to the public, nor have details been given on how it will reach the end user. It can be anticipated that it will possibly be integrated into the many products that Google has. Maybe it ends up being a super-vitaminized version of the Assistant. The truth is that if it is possible to integrate with the multiple solutions offered by the Mountain View giant, could go on to lead this battle of generative AIs.
Meta and LLaMA, their proposal to empower AI researchers
Meta could not be left out of this game. On February 24, Facebook’s parent company unveiled LLaMA (Large Language Model Meta AI)a next-generation fundamental large language model designed to help researchers advance the field of AI.
Training smaller fundamental models like LLaMA is useful because requires much less computing power and resources Make great language models for testing new approaches, validating the work of others, and exploring new use cases.
The fundamental models are trained on a large set of unlabeled data, which makes them ideal for later optimization to perform a variety of tasks, as the company pointed out in the official statement that was released when presenting this product.
“It is a family of models of different sizes (capacity) and it is designed as a foundational model, that is, a base model that can be trained on specific texts to perform specific tasks”, summarizes Di Salvo.
And he adds: “It is a model more oriented to assist the ecosystem of AI researchers, not so much to generate products. I think that is why it is published with a non-commercial license and with the focus on constituting a base on which other solutions can be tested (here what Meta gives away is its computing power to train one of these massive models)”.
META innovated in the artificial intelligence market with its LLaMA tool.
Challenges and opportunities of these technologies
Generative artificial intelligences are a great opportunity to reduce costs, facilitate the development of tasks, and reduce time.
“In the future I would expect to see a significant increase in productive output, many slow and difficult tasks will be accelerated and simplified. I’ve seen this for a long time in the companies we work with,” highlights Fredi Vivas, specialist in artificial intelligence and author of the book How do machines think? .
However, this could have a negative impact on the labor market. According to a survey conducted in the United States by Resume Builder, one in 4 companies has already replaced workers with ChatGPT, while another 30% plan to do so.
Most turn to the system to ask you to write programming code, write content, or provide customer service. This undoubtedly represents a threat to many workers.
“We should see changes in the world of work, not in terms of people out of work. Although it is also true that there is a risk in companies that are immature in terms of understanding how to correctly execute the adoption processes of these technologies. The difference between an appropriate implementation that can generate value and one that is not going to be the existence of a team with an ethical eye on the impact of this type of system”, analyzes Vivas.
Another challenge presented by these technologies is that can be used to generate false or misleading contentsuch as false news or manipulated content, with the impact that this implies in the exercise of democracy.
Artificial intelligence has replaced an estimated 25% of workers, allowing for huge savings.
Data privacy is undoubtedly another of the debate axes around these systems, since they need large amounts of data to function properly. This may raise ethical and legal issues regarding the collection, storage and use of personal information.
There is also the risk of bias and discrimination in the content generated by these AIs, since the algorithms can be trained with partial data and, therefore, generate content that reproduces stereotypes.
In short, AIs bring with them a world of possibilities, but there are also certain risks that must be taken into account to optimize their use without harming society. In this sense, it is important to promote the ethical and responsible development of these technologies, and to address the labor challenges that may arise with the advancement of these tools.
