Image source: Generated by Unbounded AI
Written by Qian Shan
Source丨51CTO technology stack
A few days ago, according to foreign media Analytics India Magazine, based on various data analysis, OpenAI's current financial situation is not good, and it may declare bankruptcy by the end of 2024. As an industry leader who created ChatGPT and set off a global boom in large models, why is OpenAI in the current situation?
01
Daily consumption of 700,000 US dollars,
The number of users is decreasing month by month
According to reports, the daily operating cost of ChatGPT is 700,000 US dollars (currently about 5.08 million RMB). All of this is coming out of the pockets of Microsoft and other investors, and if it doesn't turn a profit soon, this day-to-day cash burn will eventually be unsustainable.
In fact, OpenAI has not only been unprofitable since developing ChatGPT, but also doubled its losses to $540 million in May of this year. Microsoft's previous $10 billion investment in OpenAI may keep the company afloat for now. But if there is no further capital injection, OpenAI may have to file for bankruptcy by the end of 2024.
According to a recent report from Investopedia, in terms of OpenAI's profitability, any AI leading company, such as OpenAI, Anthropic or Inflection, is still too early for its first IPO. Furthermore, the same report states that at least 10 years of operations and $100 million in revenue are required for a successful IPO.
Another fact that cannot be ignored is that the user base of ChatGPT is also declining. When the ChatGPT website saw a drop in user numbers in June compared to May, it was asserted that it might be because students were on holiday. On the other hand, it could also be because since OpenAI released the ChatGPT API for users, people started building their own bots instead of using the original product.
This decline is not an accident. According to SimilarWeb, the number of ChatGPT users fell further in July, from 1.7 billion in June to 1.5 billion, down about 12% from June. There has been speculation on Twitter that one of the main reasons for this decline may be that most companies prohibit their employees from using ChatGPT at work.
For this situation, OpenAI is not completely unforeseen. They are pinning their hopes on paid versions to turn things around, such as people buying APIs and using GPT-4 based chatbots or other products like DALL-E2, but the financials on this are still murky, at least as far as the status quo is concerned. Achieve balance of payments.
02
Cause 1:
rival chase
OpenAI's financial situation is not very optimistic, at the same time, it also faces threats from many competitors.
Earlier, it was thought that the biggest competitor to OpenAI in the AI race would be Google or Meta, but now that Musk's X.AI is here, the unpredictability has increased.
Musk, who said he was worried about social problems caused by GPT-4, established the artificial intelligence startup X.AI to compete directly with OpenAI. There are even reports that Musk has purchased tens of thousands of GPUs from Nvidia to power his AI projects.
Musk announced the plan in an interview with Fox, and said he would create a chatbot called "TruthGPT," "or a maximally truth-seeking artificial intelligence that can understand the nature of the universe. I think This is probably the best path to (AI) safety, in the sense that an AI that cares about understanding the universe is less likely to exterminate humanity, because we are an interesting part of the universe."
Obviously, compared with ChatGPT, the concept advertised by TruthGPT pays more attention to AI security. Of course, Musk is far from the only one aiming at this point.
In 2021, the collective resignation of a group of OpenAI early/core employees has attracted considerable attention in the industry. After a period of collective silence, a new company, Anthropic, was born.
This company aims to solve the long-standing "black box" problem of neural networks, improve AI security and explainability, and has high hopes since its inception-wrestling with OpenAI. These masters who left OpenAI did not live up to expectations, and soon launched the AI assistant Claude, who gave him the nickname "GPT Killer".
In fact, over the past five years, more than 30 executives, engineers, or other employees have left OpenAI to start businesses, raising a total of more than $1 billion in financing. These people who left OpenAI, either because of different ideas or because of conflicts of interests, but no matter what their original intentions, they have emerged as a new force that cannot be underestimated: while rapidly growing into a strong opponent of OpenAI, while bringing cutting-edge technical fire Sprinkle to the farther borders.
03
Reason two:
The Threat of Open Source Megamodels
When counting the reasons why OpenAI is in trouble, the turning point that must be mentioned is the rise of open source large models. In particular Meta launched LLaMA 2 and announced a partnership with Microsoft. Opening the model to commercial and research institutions for free will undoubtedly promote the further expansion of the capabilities and coverage of open source large models.
Instead of opting for the paid, proprietary and restricted version offered by OpenAI, why don't people choose Llama 2 which is easy to modify? And in some developers' practices, it's also much better for some use cases than GPT.
In the face of the threat of open source large models, OpenAI has not remained silent. Recently, Jan Leike, a machine learning researcher and head of the alignment team at OpenAI, warned of the loss of control of open-source models.
Leike emphasized that if these self-propagating open source models are allowed to spread, it may lead to various criminal activities that are difficult to control. Who will be responsible for it will inevitably lead to disputes. In his view, if the big model and AI go wrong and are able to resist any attempt to stop it, it will really pose a threat to society as a whole.
In previous related discussions, many experts have summarized the threats and potential threats that may be caused by the misuse of AI or large models: such as audio/video impersonation, unmanned vehicles as weapons, targeted phishing, interruption of artificial intelligence control systems and Mass extortion.
But Leike's remarks have drawn a lot of criticism. Some people think that his remarks are just a strategy to hinder the development of open source models, because LLaMA 2 is stealing OpenAI's thunder.
Hugging Face co-founder and CEO Clement Delangue responded to Leike's post, noting that the tweet could be interpreted as exploiting fear to undermine open source practices.
Delangue points to uncertainty about the notion that technology "survives and spreads on its own." He goes on to ask the relevant question whether the act of creation itself poses a risk, because if a technology is really good, it should naturally find its way to widespread adoption and not just rely on open source.
04
Reason three:
Shortage of GPUs
In addition to the competitors surrounded by wolves and the direct threat of open source large models, another factor that limits OpenAI's further development is the shortage of GPUs.
OpenAI CEO Altman has said that a shortage of GPUs in the market has prevented the company from training further models and improving them. Although OpenAI has applied for a trademark for "GPT-5", it is clear that the company wants to train it, and this indirectly caused a huge drop in the output quality of ChatGPT.
Unfortunately, if OpenAI does not get more funding soon, OpenAI may have to file for bankruptcy at the end of 2024 to buy more NVIDIA GPUs that will be launched in the second quarter of this year and start training their models. At that time, competition will be more intense, losses will be more, and users will be fewer.
05
Conclusion:
Before the final chapter, how can we know life and death
Although the reasons for OpenAI's beleaguered situation are listed above, the battle for large models is far from the end, and it is too early to draw a conclusion on the future of OpenAI.
As Gary Marcus, a well-known AI scholar, commented on this analysis on Twitter: "I would like to say that I think this prediction does not take into account the possibility that software will become more efficient over time. It also does not take into account Microsoft. More money may be given to OpenAI in exchange for more control. So I would be surprised if (it goes bankrupt) in 2024. Still sobering.”
In fact, the problems faced by OpenAI, such as the shortage of GPUs, are also inevitable for its competitors. For another example, although the impact of the open-source large model on the closed-source large model is very explosive, is the closed-source necessarily inferior to the open-source? Not necessarily. Open source is the general trend, but closed-source large models still have their existence. Especially in terms of industrialization, closed-source large models may have more advantages in terms of reliability and long-term service capabilities. At least for now, the open source of GPT-3.5 only reveals the tone, and the specific progress is still unknown. It can be seen that OpenAI is far from reaching the end of the road. As for the future, we can still keep an open mind and wait and see.
Reference link:
https://analyticsindiamag.com/openai-might-go-bankrupt-by-the-end-of-2024/
https://analyticsindiamag.com/openai-raises-alarm-over-open-source-ai-dangers/
https://zhuanlan.zhihu.com/p/645626973
Babbitt Park is open for cooperation!
Chinese Twitter: https://twitter.com/8BTC_OFFICIAL
English Twitter: https://twitter.com/btcinchina
Discord community: https://discord.gg/defidao
Telegram channel: https://t.me/Mute_8btc
Telegram community: https://t.me/news_8btc
