Several days ago, Google released its new model of generative artificial intelligence called “Gemini”. Andpointed out The company’s CEO, Sundar Pichai, said that with this model, the world is on a date with entering a new era of… artificial intelligence Super powerful.
This announcement comes at a time when the company… Google It has been criticized for lagging behind Open AI, which has become the leader in generative artificial intelligence since it launched its popular model nearly a year ago.gbt chat” (ChatGPT).
Critics believe that Google, through this offering, which was accompanied by a major promotional campaign, is seeking to restore its reputation in the field of artificial intelligence and restore some of its pride, which diminished with the setbacks that accompanied its previous model, “Bard,” which has become widely described as a model. For monotony and boredom, not for artificial intelligence.
Jimni…a family model
Google plans for the Jimny to be far from single-version models, but rather a family model to which many versions belong that have varying capabilities based on use and need.
Initially, the “Jimni Nano” model, intended for operation on Android devices, is ranked lower in this family. The highest version of it is called “Jimni Pro,” which will enter service soon. It is likely to run many of Google’s artificial intelligence services, and it will also be the backbone of the “Bard” model.
As for “Gemini Ultra,” it will be the most powerful version of this family, and it is expected to be the most powerful artificial intelligence program for large languages (LLM) created by Google, and it will be largely allocated to data centers and complex applications belonging to large institutions and companies.
Currently, the basic Jimny model is limited to text commands in input and output operations, and therefore it will be too early to judge the effectiveness and efficiency of the model compared to more advanced models such as GPT4 chat. However, Google CEOs hope that soon the more advanced models in Gemini will be able to efficiently handle images, video and audio as well.
Despite Google’s assertions that it is not merely seeking to compete for the most advanced models of generative artificial intelligence, it emphasizes at the same time its endeavor to reach general artificial intelligence, which is the model that comes closest to simulating a human with its abilities to understand, comprehend, and sense, and it is a goal that Google sees. It began its first steps with the launch of the first models of the Gemini family that will “acquire more senses” and become conscious, according to Demis Hassabis, CEO of DeepMind Technologies at Google.
As usual, Google’s new launch of the Jimny family was accompanied by many doubts and questions about the true efficiency that the model might have, especially since its promotional advertisement for the product on the YouTube platform, which received more than one million and 600 thousand views, did not have the required transparency. In article In an op-ed on Bloomberg, the article’s author, Parme Olson, claimed that Google misrepresented the power of the Jimny in its promotional video. The video showed off the impressive capabilities of the Jimny model, but in reality these capabilities appear to be questionable.
AndHe appears The six-minute video showcases the Jimny’s multimedia capabilities, such as recognizing images, suggesting games, searching for balls hidden under a cup, and logical questions such as choosing the safest route. It was noted that the response process was extremely fast, which was something that surprised viewers.
But by clicking on the description attached to the video, as a disclaimer from Google, there was a sentence indicating that “for the purposes of this demonstration, latency has been reduced and the Jimny’s response has been shortened for brevity.” This objectionable remark raised the ire of many observers and critics, including Olson, who pointed out in her editorial that Google admitted that the demo video did not occur in real time using spoken commands, but rather through the use of static images and text commands that the Gemini model later responded to. Olson believes that this is completely contrary to what Google tried to suggest, that anyone can have a smooth conversation with Jimny and get the desired results in record time.
Olson asserts that Google, through the process of editing the video clip, carried out a manipulation process to convince viewers that its Gemini model has the same capabilities as the GPT Chat model, while the truth is that Gemini is still, at best, similar to the first version of GPT Chat, which Released more than a year ago.
For its part, I tried Google and… Tongue Oriol Viñales, vice president of research and learning at DeepMind and co-developer of the Gemini model, sought to mitigate both these criticisms and concerns by emphasizing that all “user commands and outputs shown in the video are real, and are Abbreviated only for the sake of brevity. He adds that the goal of the video is to illustrate what multiple user experiences created using Gemini might look like, stressing at the same time that Google did this to “inspire developers.”
Critics point out that editing operations on videos may be understandable and acceptable to a certain extent, but they raise many questions about the authenticity of the product, especially since technology companies have a long history of manipulation operations through which they aim to inflate the usefulness and effectiveness of their products. They believe that Google could have taken a more transparent and credible path by making the model available to the public for direct and actual use and obtaining real feedback.
This behavior of Google indicates that the company still suffers from an inferiority complex in the face of the amazing successes achieved by the GPT chat model, and this is understandable. Google was one of the first companies to launch the artificial intelligence race, only to lose it unexpectedly to the emerging company “Open AI”. This loss left her with a kind of loss of certainty, so she then tried to catch up.
Despite the state of emergency declared by the CEOs within the company, Google has not yet presented a convincing and attractive model of artificial intelligence to the public. Of course, this does not mean that Google is out of the race, as its model of artificial intelligence, “Alpha Code 2,” is likely to revolutionize the world of encryption on a global level.
We are in the era of the artificial intelligence race, and despite attempts to mislead here and there, the established fact is that humanity is steadily approaching achieving general artificial intelligence that matches human capabilities and even surpasses them in all fields.
