ChatGPT: The power behind AI tools changing the tech landscape and the little-known facts about it

##Since the launch of the artificial intelligence tool ChatGPT in November 2022, this new technology has demonstrated the power and impact of artificial intelligence on our lives. potential.

Sam Altman, CEO of Open AI (the company behind ChatGPT), admitted earlier this month that he was even "a little scared" of the powerful technology his company was developing. Altman predicted that artificial intelligence "will eliminate many existing jobs," but he also said that the technology will have a positive impact on mankind because it has the potential to transform industries such as education.

But who is Sam Altman? What is behind this new technology?

ChatGPT is an artificial intelligence chatbot , whose core function is to imitate human conversation. Users around the world use ChatGPT to write emails, debug computer programs, answer homework questions, play games, write stories and lyrics, and more.

Altman said in a recent interview with ABC News, "The emergence of ChatGPT will eliminate many existing jobs. This is true. However, we can make better choices, Let it have more positive impacts on our lives and further improve our lives. The reason for developing artificial intelligence is that it will be the greatest technology developed by mankind to date. The prospects of this technology, which excite me the most Yes, it's able to provide personalized learning for each student - great personalized learning."

Thanks to the huge potential of the company's technology, Altman has been in A large number of business transactions were conducted.

In January this year, OpenAI expanded its partnership with Microsoft, which will inject nearly $10 billion into OpenAI. As a result of the deal, Microsoft will likely receive a majority of the company's profits in the coming years (it invested $1 billion in OpenAI three years ago).

In addition, Microsoft is planning to apply the tool to its existing ecosystem to further improve software such as Microsoft PowerPoint, Excel and Teams.

While the influx of cash has given the company more resources, it has reportedly divided more than 300 employees and angered some in the AI ​​field. They argue that the once humanitarian company is now primarily concerned with making money.

Dr. Mike Capps is the co-founder of Diveplane, an ethical artificial intelligence company based in Raleigh, North Carolina. He served as president of Epic Games, which developed games such as Fortnite and Gears of War. He believes that OpenAI would not have been as successful without the important connection with Microsoft, but he also expressed disappointment with some of the business decisions made by ChatGPT developers.

He said, "I feel like they kind of sold their souls to speed up the process, and it turns out they did."

Business giants and AI researchers also point out that OpenAI has not delivered on its promise to make ChatGPT open source, allowing businesses and computer scientists to manipulate and customize the tool to their liking, as part of the company's increasingly "profit-centric" mentality. Another sign.

Capp added, “They kept swearing that they were going to give everything away because that was the best way to deal with the space, but now they’re not, they’re giving it away Things are torn down so you can't reproduce their work. It's very frustrating."

Dr. Chris Mattmann, chief technology and innovation officer at NASA's Jet Propulsion Laboratory, believes that OpenAI The development trajectory directly reflects the development history of the Apache Software Foundation.

Founded in 1999, Apache was originally an American non-profit company that supported open source software projects, but over time it became a "toxic place" and abandoned Many originally had altruistic intentions and instead pursued commercial interests.

Mattmann explained, "Even under elite rule, there are still political controls and committees. It works a lot like the government's dark money. It's almost like technology The concept of gray income in the industry."

While their initial goals centered around data sharing agreements and collecting data to manage, the decision to accept large donations created inherent tensions , and pushed OpenAI into a position where it had to work for the benefit of investors.

Mattmann added, "They don't release their own models. For that matter, I even appreciate Meta more for releasing Llama and allowing people to download it. OpenAI can't do that. You can't download their models. You have to pay to play the game, which is very different from what they said in the beginning."

Early this year, OpenAI has announced a waiting list for the commercial version of ChatGPT, and customers can pay to register a version of the robot that can be integrated into various products and businesses.

While many technologies have received enthusiastic attention from consumers over the past 20 years, none has received as rapid attention as ChatGPT.

"Remember how big mobile phones were? ChatGPT is much faster than that! Remember how big Twitter is? ChatGPT is faster!"

The first generation of artificial intelligence tools was launched in November 2022, and the number of users exceeded 1 million in just 5 days. By comparison, it took Netflix 41 months, Facebook 10 months, and Instagram nearly three months to reach similar metrics.

The technology’s overwhelming success has sparked countless debates about how and where to implement it, fact vs science fiction, and artificial intelligence is back in Silicon Valley after years of big promises and false starts. A hot topic in the boardroom.

Capps added, “It’s great in some ways, but never appropriate in other ways, we just have to use it correctly.”

Big companies are divided on ChatGPT. While some companies have implemented this technology to improve user experience, such as Netflix, others have banned ChatGPT entirely from their ecosystems due to lack of available knowledge and uncertainty.

ChatGPT has the potential to replace the entire business. For example, a company that created an AI to read and analyze legal documents could leverage ChatGPT, which can provide the same functionality at a lower cost.

Unlike large language models (LLMs) that are specialized in specific areas of expertise, ChatGPT is not an expert in expertise, but can provide relatively detailed responses on a wide range of topics, even if the output Prone to inaccuracies.

While the generative AI technology behind ChatGPT has been around for a few years, the OpenAI tool’s streamlined user experience and incremental improvements to the algorithm have propelled it into everyday use with mobile phones and social media use.

You ask it questions and have conversations with it, and it tries to statistically predict the best input, usually a word, sentence, or paragraph, using a large portion of all publicly available written text. Large portion. The more data fed into it, the better the AI ​​usually performs.

These forms of artificial intelligence typically use neural network-based models that assign probabilities into a large matrix of variables and filter them through a vast network of connections to produce an output.

This AI tool can generate and debug code to help build apps and websites, write emails and papers, quickly provide answers to accelerate research, and create marketing for a variety of businesses and SEO strategies and provide ideas to support creative thinking.

Capps said the project is of great significance to people whose native language is not English, people who need help writing letters, or people trying to find the best places to visit, but should not Use in situations where human health or livelihoods may be affected. For example, when your child is sick, you shouldn't ask ChatGPT how much Tylenol to give them because that would be irresponsible.

GPT3 (the company’s previous language model, launched in 2020) propelled OpenAI to new levels of popularity, using more than 175 billion statistical connections and serving in Two were trained on the Internet, including Wikipedia and numerous books. Over time, the company will improve and expand the data sets the tool is trained on.

The latest version of the tool, GPT4, was released earlier this month. OpenAI claims it can provide more information, understand and respond to images, process 8 times more words than previous versions, and is less likely to respond to malicious requests.

But ChatGPT is still essentially a “black box” in which the pedigree and source of information are not obvious. When "hallucinations" occur in code, users are unable to determine where inaccurate information comes from, which emphasizes the importance of human-driven review.

In an interview with ABC News on March 16, Altman admitted to being concerned about ChatGPT’s sometimes unreliable behavior. He said, "The thing I most want to remind people is the problem of what we call 'illusion'. This model will confidently state things as if they are completely fictional facts."

Critics also claim ChatGPT has a liberal bias, a "shortcoming" that Altman said the company is working to improve. Generative AI is susceptible to bias from many different vectors, including the user’s input, the data sets it is trained on, and the parameters and safeguards set by its developers.

Altman said in early February that the company was revamping ChatGPT’s default settings to make it “more neutral” and “empower users” to have the system reflect themselves “on a broad scale.” personal preference.

Altman told ABC News, "We are talking to various policy and security experts, auditing the system to try to resolve these issues and roll out something that we believe has good security performance. Again, we won’t get it perfect the first time, but it’s important to learn lessons and find advantages when the risks are relatively low.”

As Altman works to quell concerns about bias within his system, he has also come under scrutiny for his political donations.

In addition to hosting a fundraiser for Democratic presidential candidate Andrew Yang at his San Francisco home in late 2019, Altman has donated more than $1 million to Democrats and Democratic groups, including This includes donations of $600,000 to the Senate Majority PAC, $250,000 to American Bridges PAC, $100,000 to the Biden Victory Fund, and $150,000 to the Democratic National Committee.

In 2014, Altman co-hosted a fundraiser for the Democratic National Committee at Y Combinator’s offices in Mountain View, California, which was attended by then-President Obama.

Altman, who served as CEO of Y Combinator—an incubator startup that launched Airbnb, DoorDash, and DropBox—from 2014 to 2019, has written in multiple blog posts and talked about China in the interview. In 2017, Altman said that he "felt more comfortable discussing controversial ideas in Beijing than in San Francisco" and that he felt it was "important" to expand into China because "some of the most talented entrepreneurs" he had encountered had been Doing business in China.

In addition to this, Altman has connections with many well-known names in the technology field.

In 2015, Altman worked with Tesla and Twitter CEO Elon Musk, PayPal co-founder Peter Theil, and OpenAI, a company headquartered in San Francisco, was established with huge funding from Silicon Valley heavyweights such as LinkedIn co-founder Reid Hoffman.

At the time, the company was just a small non-profit laboratory focused on academic research, and later grew into a technology giant (valued at $29 billion) and became a major disruptor in the industry By. The company's continued advances in artificial intelligence prompted Google to sound a "red alert" internally, fearing that ChatGPT could usurp its search engine dominance.

OpenAI raised approximately $130 million from 2016 to 2019, according to a Fox News Digital review of OpenAI’s tax returns. During this period, the group invested funds in many artificial intelligence projects.

OpenAI’s tax return shows that in 2016, the company spent $10.5 million to build a research team, set goals, and select its first projects. The group also launched OpenAI Gym Beta, published nearly six comprehensive research papers, hosted a self-organized machine learning conference, developed infrastructure, and established a security team.

The next year, in 2017, OpenAI spent $28 million to demonstrate that "reinforcement learning algorithms could defeat the world in a limited version of an advanced multiplayer game called Dota 2" The best of humanity." The nonprofit also participated in a report on potentially malicious uses of artificial intelligence and published the findings, according to the tax return.

In 2018, the organization spent nearly $50 million launching the OpenAI Fellows and Scholars program. They also trained a "human-like robotic hand to manipulate physical objects with unprecedented flexibility and scaled its reinforcement learning algorithm to defeat 99.95% of teams of Dota 2 players."

In 2019, OpenAI invested nearly $2 million to create "OpenAI, L.P.," a new capped profit company designed to help rapidly scale investments in computing and talent while including checks and balances to promote the organization duty of. Tax records from that year show that through control of its partners, the group’s reinforcement learning algorithm “became the first artificial intelligence to defeat a world champion in an esports game.” These same algorithms were then used to train a pair of neural networks to solve a Rubik's Cube using human-like hands, which requires unprecedented flexibility.

According to TechCrunch, Altma’s other non-profit organization OpenResearch has received approximately $24 million in investment since its establishment.

The above is the detailed content of ChatGPT: The power behind AI tools changing the tech landscape and the little-known facts about it. For more information, please follow other related articles on the PHP Chinese website!