Revealing the secret behind the blockbuster ChatGPT: Born quickly in 13 days, the CEO forbade employees from showing off their achievements on social networks

One day in mid-November 2022, the staff at OpenAI received an unexpected task: Release an AI chatbot in 2 weeks. According to a senior director, this chatbot is named “Chat with GPT – 3.5”, and is expected to be offered for free globally.

The above announcement confused some OpenAI employees. For most of 2022, this AI company has been working to release GPT-4, an AI that is extremely good at writing essays, programming… After months of testing and tweaking, GPT-4 is almost ready. According to three OpenAI insiders, the company’s plan is to launch GPT-4 in early 2023 with some chatbots allowing users to try it out for themselves.

However, the OpenAI board of directors changed plans. Some executives fear that they will be surpassed by competitors when they launch chatbots before GPT-4 and that the early launch of an old language model like GPT-3.5 will help them receive feedback and reviews from users to improve their performance. improve new version.

So they decided to pre-launch an unreleased chatbot using an improved version of GPT-3, the language model coming out in 2020, and ChatGPT today. 13 days later, ChatGPT was born.

In the months that followed, ChatGPT became a global phenomenon. Millions of people have used this tool for poetry writing, application programming and even psychological counseling. It is also being adopted by news publishers, marketing agencies, and business leaders, sparking a new wave of investment in the next wave of AI.

At the same time, ChatGPT has also caused a lot of controversy with users complaining that ChatGPT tends to give biased or incorrect answers. Some researchers have commented that OpenAI is reckless. Many public schools in the US have banned ChatGPT so that students do not use it to help with homework.

As famous as it is, so far there is still very little information about the origin and strategy behind ChatGPT. With OpenAI internals, ChatGPT has achieved unexpected success but also brought with it many headaches.

Before ChatGPT launched, some OpenAI staff did not think the project would be successful. Previously, a similar chatbot released by Meta (Facebook’s parent company) failed after only 3 days of launch. That led the OpenAI staff to think that a chatbot built 2 years ago based on AI seemed boring.

But 2 months after launch, ChatGPT already has more than 30 million users and receives about 5 million visits per day. This number makes OpenAI the fastest growing software in history. By comparison, it took Instagram almost a year to get to its first 10 million users.

However, strong growth also comes with difficulties. ChatGPT is often overloaded with huge workloads. Meanwhile, users began to find ways to circumvent the law and violate some of this chatbot’s safety terms. The popularity of ChatGPT also caught the attention of many tech competitors. They scoffed that the underlying technology of ChatGPT is actually nothing new.

In addition, until now, ChatGPT is still an expensive investment because there are no ads. According to a post by CEO Sam Altman, each OpenAI conversation costs money to operate, and the total cost can reach millions of dollars per week. To compensate, the company has launched a version of ChatGPT Plus that charges 20 USD/month.

Despite the limitations that remain, ChatGPT’s success has catapulted OpenAI into the ranks of Silicon Valley power players. The company recently reached a $10 billion deal with Microsoft. Meanwhile, Google has declared a “red alert” and called for help two co-founders who left the company a few years ago to return to participate in AI projects to increase competition.

Altman says OpenAI’s current goal is to create what’s known as “generalized artificial intelligence” – or AGI, capable of learning human thinking.

Even so, the rapid success of ChatGPT made Altman worried that it would make users expect too much from the next versions. On his personal Twitter, the CEO tried to tone down the excitement of users, warning that ChatGPT is still limited and that relying on it to do anything important at this time is wrong.

Additionally, Altman also discourages employees from bragging about ChatGPT’s success. In December of last year, a few days after the company announced that more than 1 million people had signed up for the service, Greg Brockman, president of OpenAI, wrote on Twitter that ChatGPT had reached 2 million users. Altman later asked Brockman to delete the information, saying it was unwise to do so.

OpenAI is an unusual company, by Silicon Valley standards. Started in 2015 as a nonprofit research lab by a group of technology leaders including Altman, Peter Thiel, Reid Hoffman, and Elon Musk, it founded a subsidiary in 2019 and reached a $1 billion deal with Microsoft. Since then, the unit has grown to about 375 full-time employees, according to Altman.

In Silicon Valley, Altman is becoming the new focus. In addition to Microsoft’s $10 billion investment, he recently met top executives from Apple and Google. OpenAI also signed an agreement with news site BuzzFeed, to use its technology to create AI-generated lists and quizzes.

The race is on AI heating up globally. Baidu – the Chinese tech giant, is preparing to introduce a chatbot similar to ChatGPT in March, according to Reuters. Anthropic, an AI company founded by former OpenAI employees, is said to be in talks to raise $300 million. And Google is also racing with dozens of different AI projects.

Then there’s the GPT-4, which is expected to launch this year. At that time, it is not excluded that ChatGPT will be eclipsed, but it is possible that OpenAI’s current chatbot is too powerful so it will no longer attract users.

Source: NYT