More

    Where Is the Brink of the Future? How to Make AI Work Better

    Published on:

    If there is a will, AI will be many times better than current capabilities that end with generating text, images, and program code – Rad Okhotnikov did some simple calculations.

    Why AI isn't getting any better: Lado Okhotnikov, founder of Meta Force, on language models

    If you wake up one day and hear someone cooking in the kitchen, and at the same time you live alone, you can say that AI has achieved full-fledged intelligence and will eventually be able to perform some of your daily tasks. You will be able to delegate things to others. artificial friend.

    However, we are increasingly hearing about the potential threats posed by artificial intelligence. Concerns about AI are mostly dark, from world domination to threats to human life. But we may be overlooking its potential positive aspects.

    Artificial intelligence can be an ally rather than an enemy. In the fields of scientific research, medicine, and solving global problems, AI can process vast amounts of data and provide effective solutions. However, it is premature to call artificial intelligence as such, as it is similar to neural networks. ” To tell Rad Okhotnikovfounder of Metaforce Metaverse, where machine learning may be used.

    Rad Okhotnikov is leading the team that created the first Metaverse, a unique platform based on the principles of an entirely new business model. His views on technology development not only change the game, but rewrite the rules of making money.

    In his opinion, AI adoption is still taking too long. He argues that it is too early to fully understand how artificial intelligence will work in different situations.

    We are in the early stages of creating a metaverse, and implementing AI requires careful research. We want artificial intelligence to be able to not only cope with the tasks it is given, but also adapt to the different situations it encounters in the digital world.” says Rad.

    test

    To test how smart the AI ​​is, we asked the robot to make coffee in the kitchen. “It's not difficult!” you might say. And if we are talking about people, you would be right. But there's a problem. Kitchens come in all shapes and sizes, and coffee can be ground, bean, or instant.

    Because of this complexity, AI that isn't smart enough will be confused about how to properly brew coffee in different situations. Therefore, this test is like testing how capable artificial intelligence is to think independently and deal with different situations.

    Simple tests like this showed us that we needed to ensure that our assistants could not only follow instructions, but also make independent decisions in a variety of situations. Your cup of coffee should be as special as you like it.” commented Rad Okhotnikov, founder of Metaforce Metaverse.

    It's still hard to imagine what artificial intelligence would do in the metaverse other than writing simple code. The original idea was that here users can escape from reality, but still be able to continue communicating with people. So what's the interest in interacting with robots if human behavior is more unpredictable? It's interesting and exciting, and there are no clichés or pre-prepared answers.

    what people say.Rad Okhotnikov talks about his future profession

    On June 5th, at a certain venue, technology conferences Sam Altman and Ilya Satskeva, leading experts in innovation and artificial intelligence from Tel Aviv, spoke about the potential risks associated with using this technology.

    The issues raised by the conference participants concerned three important aspects: the impact of AI on the labor market, the threat of its use to the detriment of humanity, and the risk of a possible loss of control of artificial intelligence systems. .

    When asked about the loss of control, Ilya said:It is reckless to create an intelligence that cannot maintain control on its own” This concern is related to the possibility of the emergence of independent systems beyond the limits of human perception.

    Sam Altman agreed with his colleagues, emphasizing the importance of balancing technological advances with ensuring human control over them.

    Asked about changes in the labor market, Satskever emphasized that some jobs are already supported by AI. He noted that new jobs are certain to emerge and a period of economic instability is inevitable.

    Perhaps many professions will change their format. But we have forgotten how telephone operators, mousetraps, lamplighters, raftmen, and lecturers disappeared. I am.New industries will emerge that require the participation of people with slightly different skills.” said Rad Okhotnikov. He highlighted that an example of such an evolution is IBM's computer program Deep Blue, which defeated world chess champion Garry Kasparov in 1997.

    Because of this, many thought that after the program became the best at this game, interest in chess would fade.But the opposite happened – chess became even more popular, and the level of people's playing increased.” added Mr. Rad.

    What is the difficulty of creating advanced AI?

    Developing artificial intelligence is a complex task with many pitfalls.

    The main problem is that the development of superintelligent AI could lead to mass unemployment. Robots and computer programs will replace unskilled workers. This mainly affects people who work in menial jobs. And as unemployment increases, so do social problems.

    Moreover, scientists still do not know exactly how the human mind develops. After all, in millions of years of evolution, not a single ape ever became a human. The reason why it is impossible to set up AI development algorithms is because previous theories have turned out to be partly wrong.

    Rad Okhotnikov believes that before improving artificial intelligence, scientists need to better understand the path it will take. Otherwise, AI could do more harm than good. There are still too many unknowns in this field.

    The conclusion is

    Perhaps one of the most notable achievements to date is that of DeepMind researchers who announced the AlphaFold 2 AI model, an innovative solution that challenges the decades-old problem of protein folding.

    The new model is based on the Evoformer block, a unique architecture that combines transformers and evolution strategies. This allows AlphaFold 2 to analyze one-dimensional amino acid sequences and predict the three-dimensional structure of compounds. This, in turn, could lead to the development of new drugs and treatments for common diseases.

    A clear example of the evolution of AI.

    As for everything else, 2023 can be safely called the year of the “chatbot.” Four global IT giants have announced language models, ushering in a new era in the field of artificial intelligence.At the forefront of this innovation movement are powerful Chat-GPT 4 Language models quickly took the lead in their intelligence.

    However, if someone other than Meta Force decides to introduce AI to the Metaworld, the platform will face the still insurmountable factor of increased energy costs.

    In 2021, training a GPT-3 language model required 1.287 gigawatt-hours of power. This number was found to be significantly higher than that of previous versions. This can be explained by the increase in the number of program parameters to 175 billion. It turned out that compared to the previous model, which had only 1.5 billion functions, the new model had an order of magnitude higher energy consumption.

    There's one more problem. The number of people who play video games exceeds 2.7 billion, which is more than a third of the world's population. According to the study, computer games alone consume approximately 75 gigawatt-hours of electricity per year, which is equivalent to the energy produced by 25 average power plants.

    Adding these indicators together, the total indicator reaches several times the current energy consumption, since the metaverse processed by artificial intelligence consumes several times the current energy consumption.

    Therefore, it currently makes little sense to speculate about when language models will be integrated into the platform. “However, like many other companies, we expect to see some innovation in this space in 2024.Stay informed”, Rad Okhotnikov concluded.

    Disclaimer: This is a paid press release and BSC.News does not endorse and is not responsible for the content, accuracy, quality, advertising, products or other materials on this page. The project team purchased this advertising article for $300. Readers should conduct their own investigation before taking any action related to the Company. BSC.News shall not be responsible or liable, directly or indirectly, for any damage or loss caused or alleged to be caused by or in connection with use of or reliance on any content, goods or services mentioned in this press release. We do not owe.

    Related

    Leave a Reply

    Please enter your comment!
    Please enter your name here