月度归档 2023年4月16日

ChatGPT invested $180 million in anti-aging company, trying to increase the average life expectancy of human beings by 10 years.

In mid-2022, a startup company named Retro Biosciences came into public view, and it announced that it had obtained $180 million for a bold task: increasing the average life expectancy of human beings by 10 years. Just a short year ago, the company was headquartered in a warehouse near San Francisco. There were many containers on the concrete floor, where scientists worked and did experiments.

Retro said that as part of the "radical task", it would "stimulate the speed" and "tighten the feedback loop" in order to stop aging or even reverse it. But it is vague about the source of its funds. According to reports, it was still a "mysterious" startup, and its investors were anonymous.

Now MIT Science and Technology Review can reveal that all the investment was provided by Sam Altman, a 37-year-old senior entrepreneur and CEO of OpenAI.

Altman spent almost all his time on OpenAI, an artificial intelligence company, whose chat bots and electronic art programs have always been unique in the field of science and technology because of their comparable human capabilities.

But altman’s money is another matter. He said that he had emptied his bank account to fund two other very different but equally ambitious goals: pursuing unlimited energy and prolonging life.

He said in 2021 that one of the projects is Helion Energy, a nuclear fusion power generation startup, and he has invested more than 375 million US dollars in the company. The other is Retro, who invested a total of $180 million in the same year.

"This is a lot of money. I basically invested all my current assets in these two companies, "altman said.

Altman’s investment in Retro has never been disclosed before. This is one of the largest personal investments ever made by a startup company seeking to increase human life span.

Altman has always been a well-known figure in Silicon Valley. He used to run Y Combinator, a business incubator in San Francisco, but with the release of ChatGPT by OpenAI and its popularity all over the world, his name has been known all over the world.

It is reported that the breakthrough based on artificial intelligence technology has turned this seven-year-old company into a "member of the technology giant club". Microsoft promised to invest $10 billion in it, and altman, who has 1.5 million Twitter followers, is consolidating his status as a technology tycoon, and his invention will definitely change the society profoundly.

Altman is not on Forbes’ list of billionaires, but he is still very rich. His extensive investments include early shares in companies such as Stripe and Airbnb.

He said: "In the biggest bull market in history, I have always been an investor in technology stocks at an early stage."

Hard core technology

Now, in his words, he has invested his capital in a level that is "orders of magnitude far beyond the Y Combinator era". He has been betting these funds on several technical fields that he thinks will have the greatest positive impact on mankind: artificial intelligence, energy and anti-aging biotechnology.

Helion, headquartered in Everett, Washington, is eager to conquer atoms to create an "infinite clean energy". Joe Betts-LaCroix, CEO and co-founder of Retro, said that the company’s goal is to "prolong human life by discovering how to rejuvenate our bodies".

All these companies, including OpenAI, are what altman called "hard-core" start-ups-companies that need a lot of investment to make scientific progress and master difficult technologies. This is a shift for altman, from supporting fast-growing applications and companies in the era of Web 2.0 prosperity to supporting scientists engaged in long-term research.

Hard science companies need higher capital costs, but altman believes that their bigger goals are more likely to attract talented engineers. He recently quoted Daniel Burnham, a Victorian architect, on Twitter: "Don’t make small plans, they don’t have the magic to stimulate the blood flow."

Although nuclear fusion and life extension may be the most incredible projects (some researchers even think it’s daydreaming), few people believed that artificial intelligence would pass the medical school exam in 2023, just like ChatGPT, a chat robot of OpenAI, did this year.

Altman said that in fact, hard-core startups may have a better chance of success than simple companies. This is because there may be a thousand start-ups developing photo sharing applications, but only a few companies can build experimental nuclear fusion reactors.

enlarge the scale

Altman said that the areas he has been betting on are those where he can see some potential trends, so that technologies that seem impossible today are expected to spread quickly. This is what happened to OpenAI, which was established in 2015.

The company focuses on a machine learning technology called transformer and has steadily expanded its scale. In the process of creating its products, it only spent more than $1 billion on purchasing computing power.

The resulting model (program) can create pictures and complex text paragraphs in a few seconds, even close to human works. "We have an algorithm that we can learn, and as it performs more operations, it seems that it can be further extended," altman said.

As for the realization of controlled nuclear fusion, the trend that altman sees is bigger and stronger magnets. In order to fix the 100 million-degree thermal plasma vortex in the reactor core, we need very powerful magnets. Altman said that he initially invested about $10 million in Helion, but later increased the stakes because he was "very confident".

Although the problem of nuclear fusion has not been solved (reactors still use more energy than they produce), he has been urging Helion to list how to build several reactors every day-if nuclear fusion wants to replace coal and natural gas as the main power source, increasing production capacity is a necessary part.

"That’s what I’ve learned from my career. Scale it up and see what happens, "altman said.

Young blood

About eight years ago, altman became interested in the so-called "young blood" research. In these studies, scientists stitched young mice and old mice together so that they shared a blood system. Surprisingly, these old mice seem to have partially recovered their vitality.

This is a terrible experiment, but to some extent, it is very simple. Altman, then the head of Y Combinator incubator, asked his staff to investigate what progress anti-aging scientists were making.

He said: "It feels like this is a result I didn’t expect, and that is another result I didn’t expect. This means that something is happening quietly … Maybe there is a secret hidden here, and it is easier to find than we thought. "

In 2018, Y Combinator opened a special course for biotech companies, inviting companies with "radical anti-aging plans" to apply, but soon after, altman left Y Combinator to concentrate on his increasingly busy work in OpenAI.

By 2020, some researchers in California showed that they could achieve the effect similar to obtaining "young blood" by replacing the plasma of old mice with saline and albumin. This shows that the real problem lies in the "aging blood". Just by diluting it (and the toxins in it), we may be one step closer to drugs that resist or even cure aging.

Bates Lacroiva recalled, "Sam called me and said,’ My God, have you seen this plasma intervention paper?’ That’s not what I said, but the general idea is similar. Bates Lacroiva was a part-time biotechnology partner of Y Combinator and now leads a gathering of longevity lovers.

Bates Lacroix agrees that it’s cool and some companies should pursue it. "How about I fund you to do it?" Altman said.

But Bates Lacroix was already studying a different idea. He just finished an early joint venture project, a company called Vium. The company tried to "digitize" the rat population and added cameras and artificial intelligence to monitor the experiment. Vium has raised more than 50 million dollars, but it has not been successful. That year, it was merged into another biotechnology company, which bought its assets for $2.6 million.

Bates Lacroiva’s new plan is to set up a company to pursue cell reprogramming-another hot area involving the technology of making cells younger through genetic engineering. He has collaborated with Chinese scholar Ding Sheng Ding who has developed a new method of reprogramming cells. Bates-Lacroiva also believes that the process that cells use to process toxins (called autophagy) may be a direction worth exploring.

Altman’s answer is: "Why don’t you do all these things?"

"I will. I will set up a multi-project company around the biology of aging, which is a big project, "Bates Lacroiva recalled." Then he said,’ Great, let’s do it.’ "

The new company needs a lot of money-enough to keep it going for at least seven or eight years while conducting research, encountering setbacks and overcoming them. And it needs to finish the work quickly. The expenses of many biotech startups are decided by the board of directors, but in Retro, Bates-Lacroiva has all the decision-making power. "There is no bureaucracy here," he said. "I am a bureaucracy."

For example, Bates Lacroix did not wait for scarce laboratory space, but filled a warehouse with 40 prefabricated containers and used them as laboratories. This means that he can quickly carry out the first experiment, including repeating some plasma work on mice. Bates Lacroiva presented some preliminary results at a conference in 2022, saying that mice undergoing plasma exchange seemed to become stronger.

Mysterious start-up company

Retro employees submit experimental memos every week to record which work is progressing smoothly and which work is frustrated. Bates Lacroix said that he usually calls altman on weekends to tell him the highlights, and altman sometimes makes suggestions.

However, up to now, things related to this company in altman have been kept secret. This is a decision made by Bates Lacroiva, who wants Retro to open its own path. Altman agreed, because he tried to be "very careful not to hide the brilliance of the CEOs I worked with."

When Bates Lacroiva officially disclosed the company information in mid-2022, in a series of tweets, he still didn’t disclose altman’s name, but said that he was "lucky to get the initial capital of 180 million dollars", which was enough to "safely" support the company to run until 2030, when it planned to realize "the first proof of concept to extend life".

People familiar with the company’s thinking say it’s also because altman’s name may be a distraction. Of course, he is famous, but this reason may be wrong. Although altman’s position in the entrepreneurial world is unparalleled, his reputation is almost nonexistent in the biological laboratory and pharmaceutical industry, and a person’s scientific record is the most important in this circle.

"I’ve never heard of the name Sam altman," said Irina Conboy, a researcher at the University of California, Berkeley, although her work on plasma has amazed Sam. But she is really more familiar with Bates Lacroiva and his research on longevity.

"100 million is just a number, not a breakthrough," Comboy said.

Public discontent

Every technology also has risks. As far as artificial intelligence is concerned, it is chat bots that may spread lies and false information. As for reversing the age, if it is really effective, one of the risks often mentioned is public dissatisfaction, especially if it will be offered to the rich like altman first.

The basis of this idea is that if altman’s financial support is magnified conspicuously, Retro may become a victim and be classified as a vanity project misled by billionaires in pursuit of immortality.

We have reason to be so worried. In 2016, Peter Thiel, one of altman’s mentors, was ridiculed by the media as a "vampire looking for young victims" after he expressed interest in receiving anti-aging blood transfusion treatment. A year later, HBO’s spoof program "Silicon Valley" released an episode called "Blood Donation Boy". In the film, the CEO of a fictional technology company is in a meeting, and his vein is connected with that of a handsome young man called "blood transfusion assistant".

"We really don’t want to see those old billionaires pay for plasma donors," Bates Lacroiva said in the summer of 2022. Instead, he said, the company hopes to find more "credible" interventions, such as drugs that mimic the effect of blood replacement, for millions of people to use.

"We don’t want to treat billionaires with prejudice. I’m just saying that we don’t want to see treatments that are super expensive, awkward and difficult to implement. "

Altman said that his personal anti-aging methods include "trying to eat healthily, exercise and get enough sleep" and taking metformin. Metformin is a diabetes drug, but it is very popular in Silicon Valley because it may make people without diabetes healthier. "I hope I can use Retro therapy one day!" Altman said.

OpenAI in the field of anti-aging

Anti-aging research seems to be a promising investment field. One reason is that it has not received much funds in the past, at least relative to the scale of the research project. According to the data of the Center for Medicare and Medicaid Services in the United States-about $4.3 trillion, accounting for nearly one-fifth of the GDP of the United States-it is used for medical care, most of which is used to treat the elderly. Longevity researchers generally believe that if a drug can delay aging, it may help delay a series of serious diseases, including cancer and heart disease.

Bates Lacroiva said that in order to have the widest impact, he is looking for interventions that can scale up and affect "millions or billions" of people.

However, when Retro lifted the veil, the research on the problem of old age was becoming very popular. The Saudi government said that it will provide 1 billion dollars for related research every year, and an organization called Altos Laboratory has been established, claiming that it has 3 billion dollars. Famous investors are also involved, such as Yuri Milner, and even Jeff Bezos is rumored.

Compared with these investments, altman’s bet looks relatively small, even making Retro look insignificant. One of its projects is to test the rejuvenation technology on T cells, which are part of the immune system and play an important role in fighting infection and avoiding cancer.

These cells are particularly useful because they can be removed, rejuvenated in the laboratory and sent back to the patient. But other startups have similar goals, including Altos and NewLimit, a biotechnology company founded by cryptocurrency billionaire Brian Armstrong in 2022.

The competition for researchers is particularly fierce. Altos persuaded 24 university professors to leave their jobs and offered millions of dollars in salary and other benefits, which absorbed almost half of the top scientists in the field of reprogramming.

But Bates Lacroix also managed to attract some top talents. In 2022, he flew to Switzerland to visit Alejandro Ocampo, a researcher at Lausanne University. In 2016, Ocampo was the first scholar to study the rejuvenation of mice, and his research helped trigger the current craze for longevity investment.

"I’m glad to see Joe coming to see me in person," Ocampo said. He was so grateful for the attention he received that he later agreed to become a paid consultant for the company.

He also said that Bates Lacroiva is willing to accept his view that the age reversal of human beings will not happen in the short term. Some of Ocampo’s recent experiments explored why the reprogramming method he studied would kill some mice instead of making them live longer.

Ocampo said: "Some optimists think that we will achieve eternal life in 10 years, while some pessimists say that we will never prolong human life. As a realist, my personal opinion is that many people are choosing simple and fast experiments, but if we do the same, I don’t think we will go very far. Because this will not be an easy road. "

Ocampo said that Bates Lacroix convinced him that Retro was willing to explore these basic problems with its own funds. "They want to promote the development of science, not just pursue short-term goals," he said. "Other companies want to find an application scenario immediately, but they are willing to spend time exploring basic science."

Bates Lacroiva did not discuss Retro’s funding sources with Ocampo. Ocampo said that he didn’t know that altman was funding the startup until he was interviewed by MIT Science and Technology Review.

In an interview, altman didn’t worry about the competition from other companies. He believes that most biotechnology companies are accustomed to slow action and are usually "not good at management". He believes that the research on longevity needs an "OpenAI-style effort".

Altman said: "Retro’s main task is to become a truly excellent biological start-up company, because this is a rare thing. It combines great science with the resources of big companies in the spirit of a startup company. This is our current project. "

Original text:

The Internet is full of flawed content generated by AI. Will AI trained based on this information be outrageous?

1. Yes.

If you answer the question of "yes and no", the answer is obviously "yes".

2. The logic of AI learning can be simply summarized as the following three steps:

2.1. Input data and feature extraction: The first step of AI learning is input data and feature extraction. At this stage, the AI system will receive some input data, which can be images, text, voice or other types of data. Then, the AI system will extract some useful features from these data, which can help the AI system better understand and process the data.

2.2. Model training: The second step of AI learning is model training. At this stage, the AI system will use the input data and features extracted to train a model. This model can be neural network, decision tree, support vector machine, etc. It will learn how to map the input data to the output results according to the features extracted from the input data and features. The goal of model training is to make the model accurately predict the output results.

2.3. Model evaluation and optimization: The third step of AI learning is model evaluation and optimization. At this stage, the AI system will use some test data to evaluate the performance of the model and optimize the model according to the evaluation results. If the performance of the model is not good enough, the AI system will adjust and optimize the model to improve its accuracy and generalization ability. Generally speaking, the logic of AI learning is to continuously improve the performance and ability of AI system through input data and feature extraction, model training and model evaluation and optimization, so as to realize more accurate and intelligent prediction and decision-making.

3. Garbage input and garbage output

The quality of AI-generated content is affected by the quality of data used to train AI models. If the training packet contains defective content, then the artificial intelligence model will also be defective. This is called "garbage input, garbage output". Because the AI system will try to imitate and repeat the existing data when learning, if the data itself has problems, then the AI system may repeat these problems and even aggravate them.

Constructing high-quality data sets is the key, and attention should be paid to the source, quality, scale and diversity of data sets. Miniaturization of language model is also an important research direction.

4. But there is room for optimization and self-evolution.

4.1. The training of AI system does not only depend on the data on the Internet, and the data on the Internet is less than 5% of human information.

4.2. It also includes various artificially designed data sets and algorithms. If these data sets and algorithms are carefully designed and optimized, then the AI system can be prevented from being affected by flawed data on the Internet.

4.3. AI system can also improve itself through self-learning and adaptation, thus improving its accuracy and reliability.

4.4. Although flawed content on the Internet may have some impact on the training and development of AI system, it does not mean that AI system will become more outrageous. On the contrary, with the advancement of technology and the continuous improvement of data sets, AI systems will become more accurate and reliable.

5. Making an AIGC or ChatGPT requires a lot of technology, not just input.

Large-scale model technology accumulation: it is necessary to master the basic knowledge of large models, such as Transformer architecture, self-supervised learning, pre-training and fine-tuning.

Accumulation of natural language processing technology: you need to know the basic knowledge of natural language processing, such as word segmentation, word vector, semantic understanding, emotion analysis, entity recognition and so on.

Data set construction technology accumulation: it is necessary to build high-quality dialogue data sets to improve the quality and effect of the model. The construction of data set needs to consider many factors, such as data source, data quality, data scale, data diversity and so on.

Algorithms and computing power: You need to master reinforcement learning, generative model, attention mechanism and other algorithms, and have enough computing resources to train and optimize the model.

6. Take the manuscript to make an inappropriate analogy.

Copying refers to copying, pasting, modifying and deleting other people’s original articles without authorization, which makes them look different from the original, but in essence they copy the contents and ideas of the original. Editors usually aim to get the content quickly, save time and energy, so as to achieve the purpose of publishing articles quickly, but this behavior seriously infringes on the intellectual property rights of the original author, and also violates academic ethics and professional ethics. Washing manuscripts not only harms the interests of original authors, but also greatly damages the reputation and image of the whole industry, so it is regarded as an immoral and illegal behavior.

However, the manuscript washing should also be level and creative. The awesome manuscript washing often needs to "see" a lot of materials, which is another process from quantitative change to qualitative change.

7. Going back to the nature of AI, is it a tool or decision logic? What do you do with AI?

The process of human receiving information can be divided into the following stages:

Perception: Perception means that we receive external information through sensory organs, including vision, hearing, touch, taste and smell. Perception process is based on the interaction between sensory organs and external stimuli, which transforms external information into neural signals and transmits them to the brain.

Attention: Attention refers to the process of selecting and processing the perceived information. Because of the diversity and complexity of external information, we can’t handle all the information at the same time, so we need to filter out important information and deal with it through attention.

Understanding: Understanding refers to the process of interpreting and understanding the information we receive. This process needs to rely on our knowledge, experience and language ability to connect and integrate the perceived information with the existing knowledge, thus forming new cognition and understanding.

Memory: Memory refers to the process of storing and processing the received information. Memory can be divided into short-term memory and long-term memory. The former is the ability to temporarily store information for processing, while the latter is the ability to permanently store information in the brain.

Judgment: Judgment refers to the process of evaluating and judging the information we receive. This process needs to rely on our values, beliefs and cognitive abilities, and compare and evaluate information with our existing cognitive and value systems, thus forming our attitudes and views on information.

Action: Action refers to the process of making decisions and taking actions according to the information we receive. This process needs to rely on our willpower, decision-making ability and action ability, and turn our understanding and judgment of information into actual actions and decisions.

What is the purpose of obtaining information with AI? To what stage can AI replace you?

The brand-new Chongqing team appeared in China football! The warm-up match defeated the newly promoted Super League, which made the fans sit up and take notice.

A few days ago, Chongqing football ushered in a brand-new professional league team, which is Chongqing Tonglianglong, which gathered many old Lifan generals. After the dissolution of the Chongqing team, Chongqing Tonglianglong began to play in the China Champions League and began to carry the banner of reviving Chongqing football. Chongqing Tonglianglong performed well in the Champions League last season and got the qualification of Chong B. After Chongqing Tongliang Dragon rushed to China B, it filled the gap that there was no professional team in Chongqing football.

Chongqing lifan used to be a strong football team in China, and it also performed well in the Super League, and once beat Evergrande. Chongqing team has also trained international players like Zhang Chiming and Feng Jin. Unfortunately, before the Super League started last season, the team once again encountered the crisis of unpaid wages. Finally, after comprehensive consideration, Chongqing team decided to quit professional football. Chongqing team quit professional football, which made Chongqing football lose its professional team.

At the critical moment, Chongqing Tonglianglong stood up, started to play in the China Champions League in a team, and recruited the Chongqing football flag Wu Qing. Chongqing Tongliang Dragon has now rushed to China B, attracting the attention of a group of old Lifan generals. Chongqing Tonglianglong can consider introducing Liu Huan whose contract with Guoan expired. After leaving Guoan, Liu Huan went to Dalian for trial training.

Whether Liu Huan can stay in Dalian is a huge unknown. Because Liu Huan’s Dalian team is in a precarious state. Dalian team doesn’t even have the qualification to introduce new aid. If this problem is not solved in the future, Liu Huan will not be able to complete the registration. Therefore, Liu Huan can consider returning to Chongqing. At the same time, Chongqing Tonglianglong chose to play a warm-up match in order to test the training results during this period.

Chongqing Tonglianglong chose to play with Qingdao Manatee, a new Chinese Super League army. Originally, fans thought that Chongqing team would be played by Manatee. Unexpectedly, the Chongqing team was very tenacious in the game and finally defeated its opponent 1-0. The strong play of Chongqing team in the warm-up match made the fans see the hope of rushing to armour. I look forward to Chongqing Tonglianglong’s steady and steady progress in the Chinese B League and achieving the goal of rushing to the first division in the new season.

Inzaghi faced two fateful battles at Inter Milan and Kong Di’s three choices in Italy.

Inter Milan lost to Spezia 1-2 away, which put coach Inzaghi in a delicate position at Inter Milan.

Next, the Nerazzurri need to face Porto in the Champions League and Juventus in Serie A..

These two games will play a decisive role in Inzaghi’s future.

Inter Milan’s minimum requirement for Inzaghi is to keep the top four position in Serie A, otherwise the coach may face dismissal after this season.

People’s doubts about Inzaghi are increasing.

On the other hand, Kong Di and Tottenham are both bored, and they will probably leave Tottenham after this season, and may even be fired in advance.

If Kong Di does not choose to coach in other leagues this summer and returns to Italy, then he has three main options.

At present, Kong Di’s net annual salary at Tottenham is about 10 million euros, which is one of the biggest obstacles for Kong Di to coach in Serie A..

Juventus and Inter Milan are two potential destinations for Kong Di in Serie A..

For Juventus, the current coach allegri’s net annual salary is about 7 million euros, and his contract with the team expires in the summer of 2025, which means that Juventus will have to pay a high price for firing allegri and signing Kong Di.

In addition, it is also possible for Kong Di to return to Italy for a year and then return to work.

These are Kong Di’s three main choices in Italy. Of course, there may be teams outside Italy who will hire Kong Di to coach.

Paris Saint-Germain is willing to spend 50 million pounds to introduce Maguire?

Paris Saint-Germain is preparing to offer 50 million pounds for Harry Maguire in the summer and bring him to France..

It is reported that Manchester United are willing to break the club record and sign Eduardo Camavinga from Real Madrid for 115 million pounds..

Tottenham Hotspur is facing a struggle to convince Thomas Tuchel to become their new head coach. It is reported that both he and Mauricio Pochettino are the best candidates to succeed Antonio Conte, and Chairman Daniel Levy has reservations about bringing back former head coach Pochettino..

Manchester United has added Mason Mount of Chelsea to the list of attacking midfield targets including ude Bellingham. If Erik ten Hag spends most of his summer budget on bringing in Harry Kane or Victor Osimhen, the summer transfer window will exceed the budget..

The exit of Paris Saint-Germain in the Champions League is likely to prompt the departure of head coaches Christophe Galtier and Sports director Luis Campos..

Considering the cost of living crisis, the soaring business income of Chelsea, the performance of the club this season and the possible impact of the rising ticket fees on the atmosphere of the match day at Stamford Bridge, Chelsea fans warned the club president Todd Boehly that if he decided to raise the ticket price after a bad season at Stamford Bridge, it would cause "irreversible harm" to the fans..

Before Saturday’s game against Brighton, Leeds may get unexpected good news, and striker Rodrigo may play only one month after ankle surgery..

Wolves hope to beat Premier League rivals Tottenham Hotspur and West Ham United and plan to sign Alex Scott of Bristol City for 20 million pounds..

Daniel Levy, chairman of Tottenham Hotspur, will consider a series of candidates to change coaches, including Mauricio Pochettino, who may return to the club, and Luis Enrique, who was fired by Spain after the World Cup. Celtics coach Ange Postecoglou, Brighton coach Roberto De Zerbi and Neapolitan coach Luciano Spalletti..

[Two sessions for 30 seconds] Zhang Boli: Transforming traditional pharmaceutical industries with artificial intelligence.

[Two sessions for 30 seconds] Zhang Boli: Transforming traditional pharmaceutical industries with artificial intelligence.

On March 9th, the Tianjin delegation of the First Session of the 14th National People’s Congress held a group meeting to review the work reports of the Supreme People’s Court and the Supreme People’s Procuratorate. Representatives from Chen Miner, Qin Gang, Zhang Gong, Yu Yunlin and Duan Chunhua attended.

The beauty created by artificial intelligence AI is lifelike. These 30 beautiful pictures are amazing!

The beauty created by artificial intelligence AI is lifelike and amazing!

The beauty created by artificial intelligence AI is really amazing!

The beauty created by artificial intelligence AI is vivid and beautiful!

The beauty created by artificial intelligence AI is so real and amazing!

The beauty drawn by artificial intelligence AI is vivid and amazing!

The beauty drawn by artificial intelligence AI is real and beautiful! It’s amazing!

The beauty drawn by artificial intelligence AI is amazing! Amazing time!

The beauty drawn by artificial intelligence AI is fresh and refined, so amazing!

This may be the appearance of a small jasper, which is the opposite of the image of a good family! ! ! !

Mature, steady, square, atmospheric, it seems that this is the kind of woman!

There are beautiful women in the north, who laugh and pour people into the city, and then laugh and pour people into the country. This is probably the kind of beauty they are talking about!

Clear water produces hibiscus, which is naturally carved.

Clear water produces hibiscus, and it is natural to carve it! Small jasper, good family! Which is better?

Good-looking skins are the same, and there is no interesting soul Wan Li!

Is the skin beautiful or the soul interesting?

Welcome everyone to comment! ! ! ! !

Laundry technology sharing: can ChatGPT make autonomous driving faster?

Recently, the hottest topic in the science and technology circle is "ChatGPT". However, ChatGPT is only an external manifestation, and what deserves more attention is the development of AI technology behind it and its future application.

Some people even describe the changes brought by ChatGPT optimistically: Before ChatGPT, AI was only a module of existing scene products at most. Then, after ChatGPT, AI will redefine the product framework of existing scenarios.

Whether it is as optimists say remains to be seen, but whether autonomous driving, as one of the important scenes of AI landing, will have further development in this wave has still aroused many people’s discussion.

Some people think that autonomous driving needs more graphics, images and data processing ability, requires higher image algorithm, and has little correlation with natural language processing ability. It is not possible to realize autonomous driving with ChatGPT’s ability at present.

Of course, some people think that the appearance of ChatGPT shows us a possibility, that is, trained AI will make high-level autonomous driving expected to appear in a few years.

Why does the progress of AI technology make people pay attention to whether autonomous driving is affected?

Observing the development history of autonomous driving, it is not difficult to find that every major breakthrough of autonomous driving is synchronized with the development of AI technology.

We know that,AI is actually imitating the brain neural network and learning some very humanized skills by analyzing a large amount of data.In 1980s, the first practical application of neural network happened in the field of automatic driving.

In 1987, researchers at Carnegie Mellon Artificial Intelligence Laboratory tried to make a truck that could drive automatically. They manually write codes for all driving behaviors and write as detailed instructions as possible for various situations encountered by trucks on the road, so as to make the vehicles run automatically. But unfortunately, this way can only make the car achieve a speed of several inches per second.

Manual code writing failed, and another doctoral student named Dean pomerleau chose another way: neural network.

He named his system ALVINN. After adopting this system, trucks use the images taken by the roof camera to track what drivers are doing, so as to observe how to learn to drive on the road. In 1991, ALVINN drove from Pittsburgh to Erie, Pennsylvania at a speed of nearly 60 miles per hour.

However, a more direct and broader impact occurred in 2012.

Jeff Hinton, a professor at the University of Toronto, and two of his students, Alex Krzyzewski and Ilya Satsky, won the first prize in the ImageNet image recognition competition, and published a paper introducing the algorithm AlexNet. This paper is not only the turning point of artificial intelligence, but also the turning point of global technology industry.

As the key technology of autonomous driving, target detection and image recognition are highly benefited from the breakthrough of computer vision algorithm. Therefore, with the recognition accuracy of Li Feifei team, director of Stanford Artificial Intelligence Laboratory, surpassing humans for the first time on ImageNet open data set in 2015, autonomous driving, as one of the most important landing scenes of AI, has also entered the fast lane of development.

So, will the appearance of ChatGPT become the Milestone of autonomous driving again?

Generally speaking, AI can be divided into three parts: voice, vision and natural language understanding.The last wave of AI was mainly based on the breakthrough of visual image recognition technology, and this time ChatGPT is a natural language processing technology based on GPT-3 model, which can effectively simulate human language understanding ability, thus helping people better understand and analyze natural language text data.

When we want to discuss what impact ChatGPT will have on autopilot, we think that we should first find out whether autopilot here refers to mass-produced low-level autopilot (assisted driving) or high-level L4 autopilot. Secondly, does ChatGPT refer to a language model or a more generalized generation model?

From the perspective of natural language understanding, ChatGPT has a more direct impact on human-computer interaction in the assisted driving part, but it may not have a great impact on L4-level automatic driving.

Cui Dongshu, secretary-general of the Federation, also wrote in his WeChat WeChat official account that the innovation of human-computer interaction and intelligent cockpit system is very strong, especially the human-computer interaction ability of domestic car companies is very strong. Only China enterprises can understand Chinese more deeply. With the empowerment of the bottom layer in the future, the domestic automobile industry will have more good human-computer interaction effects at the application level.

For example, by using ChatGPT, the vehicle can interact with the driver by voice or text, and provide the driver with real-time feedback on vehicle status and driving information.

Before this, although a large number of in-vehicle interactive systems have appeared, the pain point of the industry mainly focuses on the "understanding" part, and most of the in-vehicle voice interactive systems are not intelligent in "understanding", resulting in a single function and command word of the whole system. ChatGPT’s explosion made the market see the hope of solution.

However, Cui Dongshu, secretary-general of the Federation, also said that,Electrification is the core of new energy vehicles, and intelligence is just icing on the cake. In the future, the core competitiveness of car companies will still be to build electric vehicles, and at the same time make full use of intelligence such as ChatGPT to empower the development of the automobile industry.

Of course, whether it is the core or not, it is not enough to have a technological breakthrough if you want to get on ChatGPT. An AI industry person told Titanium Media that "there are still cost issues, including the use cost, cloud service cost and targeted training cost."

However, from a broader generative model, the generative model with big data and large parameters will help to achieve a higher level of autonomous driving.

He Xiang, a data intelligence scientist at Mimo Zhixing, said in an interview with Titanium Media App that the vehicle-side capabilities mainly include two categories: perception and cognition. The perception ability really relies mainly on image technology, while the cognitive ability relies more on similar generation technology of ChatGPT.

That is to say,ChatGPT’s revolutionary significance lies in: letting AI model enter the era of knowledge and reasoning. At present, the biggest shortcoming of autonomous driving lies in the lack of sufficient intelligence in decision-making planning.

ChatGPT uses a training method called "Human Feedback Reinforcement Learning (RLHF)", and He Xiang, a data intelligence scientist at Mimo Zhixing, explained to Titanium Media APP.GPT is a large-scale universal pre-training language model. GPT1, GPT 2 and GPT 3 mainly improve the parameter scale, while ChatGPT mainly introduces human feedback data for reinforcement learning.

The introduction of this method can ensure the minimum output of useless, distorted or biased information according to human feedback in training.

It happens that there is also a kind of automatic driving decision algorithm called imitation learning, which is to let the machine learn how human drivers do it in different scenarios.

Generally speaking, every takeover by a human driver is an artificial feedback to the autonomous driving strategy; This takeover data can be simply used as a negative sample, which is a record that the autopilot decision is corrected. At the same time, it can also be used as a positive sample to improve cognitive decision-making

"The big model of big data and big parameters can learn more potential knowledge, including different environments and different scenarios, which is equivalent to learning a lot of common sense of autonomous driving. This common sense is very important for autonomous driving decisions." He Xiang, a data intelligence scientist at the end of the year, told Titanium Media App.

That is to say,In the process of autonomous driving research and development, the idea of human feedback reinforcement learning can be used to train a model to verify and evaluate the output of the machine model, so that it can make continuous progress and finally reach the driving level of human beings.

Therefore, it can be said that the improvement of basic ability has brought about the expansion of imagination and applicable scenarios. However, at this stage, we still can’t accurately judge how much change the big model represented by ChatGPT will bring to autonomous driving. An industry person told Titanium Media App that the excellent generalization ability trained by the big model may make there no corner case in the world.

Corner case refers to a small probability event that may occur during driving, but the frequency is extremely low. Although it is rarely encountered at ordinary times, it is likely to lead to a fatal traffic accident when encountering a corner case that cannot make a decision for an autonomous driving system.

The emergence of ChatGPT has made the industry realize that it is possible to gain a higher level of autonomous driving technology by constantly accumulating kilometers and running like this.

In fact, before this, both foreign Tesla and domestic Tucki, Baidu and Mimo Zhixing were already exploring the route of "big model".

In 2020, Tesla announced that it would introduce a large model based on deep neural network into its autonomous driving, and now it has realized a large-scale public beta of pure visual FSD Beta; Tucki expressed the viewpoint of using large models to get through the whole scene of XNGP on the 1024 Science and Technology Day in 2022. Baidu Apollo believes that the Wenxin model will be the core driving force of the elevator’s automatic driving ability.

As early as 2021, Mimo Zhixing announced that it would improve its data processing ability with the help of a large model. On February 17th this year, Mimo Zhixing officially upgraded the large model of human driving self-monitoring cognition to "DriveGPT", which will continue to introduce large-scale real takeover data, and continuously improve the evaluation effect through intensive learning of human driving data feedback. At the same time, it also uses DriveGPT as a cloud evaluation model to evaluate the driving effect of small models at the vehicle end.

However,The development of high-level self-driving cars is a complex multidisciplinary field, involving a wide range of technical and regulatory challenges. The progress of artificial intelligence technology can bring some impetus, but this is not a short-term problem.

It is reported that GPT3.0 involves 170 billion parameters, with more than 300 GB of memory, and the training process costs more than 12 million US dollars. The above-mentioned industry insiders said that the autopilot algorithm is to run in the car. Can such a large model be deployed to the car? How much computing power does it need to support? In addition, autonomous driving can not be completed by repetitive and simple road data stacking, so how to ensure a large amount of data is also a key issue.

The article is reproduced from the titanium media. The opinions in the article are only for sharing and communication, and do not represent the position of WeChat official account. If copyright issues are involved, please let us know and we will deal with them in time.

About the laundry list technology company

Laundry Technology Co., Ltd. is the world’s leading spatio-temporal intelligent big data service company and the leading location data and industry application solution provider in China. Relying on self-developed patented technology, based on high-precision maps and multi-source spatio-temporal intelligent big data, we will establish city-level and industry-level holographic spatio-temporal digital twin systems, and actively serve smart transportation (autonomous driving, smart highway, vehicle-road coordination), natural resource asset management (carbon neutral and environmental protection remote sensing data service), LBS smart industry applications (mobile Internet local business location service, smart travel, smart logistics, new infrastructure, smart city, emergency rescue) and other fields. For more information, please search for "laundry list technology" and visit the company website.