Have you ever seen a photo of a teapot in the form of a lawyer or read an intelligent article that deviates slightly from the subject? If this is the case, you may have discovered the latest trend in artificial intelligence (AI).
Dall-E, GPT and learning systems of the PALM machine are a name as innovative tools that are capable of accomplishing creative tasks.
These systems are classified as “foundation models” and are not all of media and celebration tips. So how does this new AI approach work? Does this mean that human creativity will find its end and a deep nightmare will begin?
1. What is a foundation model?
The foundation models work by forming a single large general information database, then adapting the diagram to new challenges. The previous models tended to start from scratch for each new challenge. To compare the photographs (like a snapshot of a pet cat) with legend (“Mr. Fuzzyboots Le Cat Tabby relaxes in the sun”) required to scan hundreds of millions of examples.
After his training, this model is able to say what cats (and other things) look like in the photos. The model can also be used for several other useful AI tasks, such as the creation of new images from a single legend (“Show me a koala that dips a basketball”) or the editing of images based on written instructions (“Make it look like this pay pay”).
2. How does it work?
Foundation models are based on “deep neural networks”, which vaguely inspire the functioning of the brain. This implies sophisticated mathematics and a considerable quantity of computing power, but it comes down to a complicated form of correspondence of motifs.
For example, a deep neural network can associate the word “cat” with pixel patterns which often appear in cat images, such as soft and blurred and hairy texture spots. The more examples there are than the model sees (the more appropriate results are shown), the larger the model (the more “layers” or “depth”), the more these models and correlations are complicated.
In a way, the foundation models are only an extension of the “Deep Learning” models which have dominated research on AI in the last decade. However, they have non -programmed or “emerging” behaviors that can be both surprising and new.
For example, Google's Palm language system seems to be able to provide explanations for difficult metaphors and jokes. This goes beyond simply imitating the types of information that it was originally designed to process.
3. For the moment, access is limited.
The scale of these AI systems is overwhelming to consider. Palm has 540 billion parameters, which means that even if everyone on the planet memorized 50 numbers, we still would not have enough storage to reproduce the model.
The models are so important that training requires significant amounts of IT resources and other. An estimate has put the cost of teaching the Linguistic Model of Openai GPT-3 to around 5 million dollars.
Consequently, only large technological companies such as Openai, Google and Baidu can afford to build foundation models for the moment. These companies limit that can use services, which has economic meaning. The limits of use can give us a certain hope that these systems will not be used for harmful purposes (such as the creation of false news or defamatory equipment) of any time. However, independent researchers are not able to question these models either and report their results in a transparent and responsible manner. We therefore do not yet know all the implications of their use.
4. What will these models bring to “creative” industries?
In the near future, more foundation models will be produced. Smaller models are already published in open source versions. Software companies are starting to experiment with the license and marketing of these services, while AI researchers work hard to make the software more efficient and accessible.
The remarkable creativity demonstrated by Palm and Dall-E 2 indicates that the creative professions could be affected by this technology earlier than expected.
As he says, the robots would first take the jobs of “blue necklace”. The professions which require creativity and education, called jobs of “white necklace”, were supposed to be relatively sheltered from automation.
However, the models of in -depth learning AI already excellent in tasks such as X -ray analysis and determination of macular degeneration of ocular condition. Foundation models could soon offer cheap and “fairly good” creativity in areas such as advertising, writing, stock illustration or graphic design.
The future of creative jobs can be a little different from what we expected.
5. What does that mean for legal facts, news and the media?
As we will not be able to say that creative content is the result of human activity, foundation models will ultimately influence legislation in fields such as intellectual property and evidence.
We will also have to face the disinformation and disinformation generated by these applications. We must already manage many disinformation problems, as we see in the Russian Invasion during Ukraine and the emerging number of Fake Fake Images and Video. Foundation models are about to revive these challenges.
It's time to plan!
As researchers who study the effects of AI on society, we believe that foundation models will cause major transformations. They are closely controlled (for the moment), so we can have a little time to consider their implications before becoming a big problem. The genius is not yet completely out of the bottle, but the foundation models are a large bottle, and inside, there is a very intelligent genius.
