Infinix’s range has been expanded with the INBook X2 Slim laptop, which, as you might guess from the name, stands out for its thin body.

Indeed, the thickness here is only 14.8 mm, and the novelty weighs only 1.24 kg. And this at a starting price of just $340!

The conditional payment for such characteristics is the platform. The laptop relies on far from the latest 11th generation Intel Core processors, that is, Tiger Lake. In general, they are quite good today, but here it is worth noting that the basic version of the laptop has a Core i3-1154G4, and this is a dual-core model. More expensive modifications already have quad-core Core i5 and Core i7, but then you have to pay a minimum of $ 475.

As for the rest of the parameters, the mobile PC has from 8 to 16 GB of RAM, an SSD from 256 GB to 1 TB, a 15-inch screen with a narrow frame and Full HD resolution, Wi-Fi 5 and 65-watt charging. The set of ports is represented by a pair of USB 3.0, USB-C and an SD card slot.

Motorola yesterday introduced foldable smartphones Moto Razr 40 and Moto Razr 40 Ultra, which turned out to be cheaper than all competitors, and the younger model even offers a revolutionary price for this segment at $560. However, these are Chinese prices, and in the global market everything is different, since new items are significantly more expensive there.

To start with, the Moto Razr 40 is called the Razr in the global market, and the Moto Razr 40 Ultra is called the Razr Plus.

In the US, apparently, only the latter will be sold at a price of $1,000. In Europe, they will already ask for 1200 euros for it. The base model Razr will be available in select countries in Europe, Latin America, and Asia, but pricing has not yet been announced. It is expected that it will be around 700-800 euros. It’s still cheaper than other similar smartphones, but not as revolutionary as $560.

One of the problems of modern neural networks is their “hallucinations”. This is the name given to data that does not correspond to the real state of affairs, but neural networks generate them along with true facts. And now OpenAI announced that it has come up with a new method for training artificial intelligence models that will help to cope with the “hallucinations” of AI.

“Even the most modern models are prone to false conclusions – they tend to invent facts in moments of uncertainty. These hallucinations are especially critical in areas that require multi-step reasoning, since one logical error is enough to make the whole result wrong.”, write the OpenAI researchers in the report.

The idea of ​​the new method is to change the motivation system for AI. In other words, it is proposed to reward the neural network for each correct step of reasoning, instead of the reward being given for the final correct answer. This will allow the neural network to be more accurate and check the facts.

According to the researchers, the detection and elimination of logical errors in the model, or hallucinations, is an important step towards creating a consistent AGI. [искусственного интеллекта общего назначения]”said Karl Cobbe, staff mathematician at OpenAI. He also clarified that the company has already opened access to the accompanying data set of 800,000 labels that were used to train the neural network as part of the study.

At the same time, the problem is far from being solved, and experts say that the company has not yet opened all the information on this study. Thus, the risk of AI “hallucinations” remains relevant for the time being.