The artificial intelligence is taking leaps and bounds, making little by little is much more implemented in society.
The clearest example can be DALL-E which has made many people interested in this kind of technologies based on GPT-3. But in addition to allowing people to create images through a phrase, it can also be interesting to have a artificial intelligence that is qualified to solve complex problems.
The math it is a science that requires a lot of logic to be able to develop its problems. In class, for example, students must understand the statement, specify the data, apply the correct formula, among other key points. a few days ago, Google has introduced an AI to be able to Solve these math questions and TechMarkup brings what it is about and how this new can be used technology.
Math problems could stop being a headache
Google announced on June 30 its new artificial intelligence called Minerva. This ad demonstrates that he is capable of solve math and science questions using step-by-step reasoning. All numerical calculations and symbols used They do not require a calculator in any case.
Below is a clear example of a problem that a student may have in front of him in his math class. Minerva can now perform the entire calculation step by step and not directly, creating the different notations and finely clearing the unknown and what is required. All this with only an initial question and without offering additional data.
Obviously this is a natural evolution of the Applications or web pages that are used to solve mathematical equations.
Now with a simple statement the resolution will be obtained thanks to being based on PaLM (Parthaways Language Model), along with 118 GB of scientific articles to preserve the most used mathematical notation by all.
But Google has not wanted to stay only in the most basic mathematical problems that exist. He wanted to take a step and go much further. That is why he has faced Minerva with a college level problem collection with a wide variety of sciences such as chemistry, engineering or biology. In all cases it has been seen that it offers great results:
The model is not perfect and also makes mistakes
As detailed by Google, Minerva is wrong from time to time, although its errors are “easily interpretable”. In the words of the researchers, “approximately half are calculation errors, and the other half are reasoning errors, in which the solution steps do not follow a logical chain of thought.
Another option is for the model to get the correct answer with faulty reasoning (false positive). Below TechMarkup leaves a couple of examples:
Finally, the researchers point out that the model has some limitationslike what model responses cannot be automatically verified.
The reason is that Minerva generates responses using natural language and mathematical expressions LaTeX, “without an explicit underlying mathematical structure”.
Google’s AI that allows you to create images from text
Google has also developed Pathways Autoregressive Text-to-Image model, Parti for its acronym in English, an autoregressive system that creates photorealistic graphic designs from text.
part is that is able to receive long and complex instructions, so its developers made an effort to make this Artificial Intelligence understand the different events that have occurred throughout the history of humanity and manifest a “precise knowledge of the world” in the words of its creators.
This latest artificial intelligence from Google was created by means of big data sequences of images and graphic content, as opposed to the typical data processing of text, numbers and symbols.
All these procedures are known as “token sequences” and are carried out through systems that are called Tokenizers, for the case of Parti, the ViT-VQGAN was used.
Finally, it should be noted that Parti can generate images with based on descriptions up to 67 words of extension.