Is Sora stuck in the research quagmire? OpenAI loses 5 billion in a year, Goldman Sachs report mistakenly triggers AI stock earthquake!

Is Sora stuck in a research dilemma?

Since its release in February this year, Sora is still in a "futures" state. Why hasn't it been opened up yet?

Recently, foreign media CNBC revealed that the reason why OpenAI's Sora model has been delayed for so long is due to "dialogue with policymakers."

As early as the end of February, Bloomberg reported that considering this year's U.S. election, OpenAI hopes to avoid security risks.

Another reason is that at that time, Sora was not technically ready for release.

The complex reasons behind the delayed "futures"

Since Sora's debut, successors have emerged in large numbers, but it has never been fully open.

The official OpenAI account occasionally releases videos created by artists using Sora.

In the latest video, a Singaporean artist, Niceaunties, tried a project with a theme of "aging, beauty, freedom, and fun" using Sora.

Niceaunties named it "Auntie's Eggs."

Another new video was created by David Sheldrick, a Korean artist living in London, showcasing a fashion show using Sora.

After watching, netizens were amazed. Some even described it as the birth of an art movement.

For many, they can only look on from afar, as OpenAI Sora is still just a tool in the hands of artists.

It is obvious that in the field of AI video models, models such as Runway Gen-3 Alpha, Dream Machine, and domestically popular video models not only approach the strength of Sora but can also be used for freeEven Midjourney is preparing to announce the AI ​​video model.

Why is OpenAI delaying the release?

Foreign media reports that Sora may be facing research difficulties. On the one hand, it is necessary to discuss security risks with the government in depth; on the other hand, it urgently needs the cooperation of Hollywood and artists.

Some netizens speculate that Sora must undergo review before it can be released.

On the other hand, OpenAI trained the model with YouTube data, facing potential legal issues, which is also a factor.

In March, OpenAI's Chief Technology Officer Mira Murati, in an interview with Bloomberg, was vague about Sora's use of YouTube data.

She only said, "We use publicly available data, as well as authorized and licensed data."

Subsequently, YouTube's Chief Executive Officer Neal Mohan warned of this possibility in an interview.

Murati has stated that it is certain that Sora will be released this year. However, the cost of generation needs to be reduced first. Currently, Sora is much more expensive than other existing AI systems.

Although not public, artists have been getting trial use

Since its release, Hollywood artists who have been able to get access to the beta version of Sora have already shown the direction that OpenAI wants to focus on in the future.

In March, OpenAI first released videos generated by Sora by seven artists.

For example, the inspirational short film "Balloon Man" that impressed many people.

In May, screenwriter and director Paul Trillo released the first official MV produced using Sora.

To create this 4-minute video, he worked for 6 weeks, generating a total of up to 700 clips, and finally selected 55 finished clips from them.

Not only targeting Hollywood, OpenAI is also "promoting" Sora to film production companies, media executives, and talent agencies, hoping that filmmakers will apply it in their work.

Director Ashton Kutcher, who tested the Sora model, expects it to have a profound impact on the film industry.

In the latest report, OpenAI is collaborating with the New York City art museum to provide artists with access to Sora for participation in the Strada gallery exhibition.

It is reported that the Strada Nuova: New Road art exhibition will last for three weeks.

Many artists, researchers, scholars, and creative individuals at the intersection of physics and digital art will come to exhibit.

Curator Paul Hill stated that about 6 months ago, he began collaborating with OpenAI to provide artists with accessible AI tools.

These include not only the Sora video generator, but also the Voice Engine speech generator, Dall-E 3 image generator, ChatGPT, as well as educational resources and artist grants.

Interdisciplinary artist Minne Atairu, who has been exploring the application of AI tools in art for the past 4 years.

She uses AI to generate 2D/3D images, videos, and other works, focusing on under-researched areas in Black history archives.

In this exhibition, Atairu used Sora to generate the video "Regina Gloriana," inspired by a supernatural horror film made in Nigeria in the 1990s.

![](https://wpimg-wscn.awtmt.com/abb575d6-519c-49b6-9ad0-58b4727a6e55.png?The developers of Sora believe that the current level of the model is equivalent to the early stage of GPT-1. Just like how GPT-1 proved that language models can scale, Sora may gradually gain new and unexpected capabilities similar to the GPT series.

Therefore, further scaling of video-generating AI models may bring revolutionary new applications.

However, it is worth noting that GPT-1 appeared very early, but it was not until GPT-3 that it truly embarked on a commercial path.

Goldman Sachs just retracted the "AI Bubble" statement, but misinterpreted the "ChatGPT collapse"

In addition to the endlessly delayed Sora, OpenAI recently had other bad news.

A report from Goldman Sachs misinterpreted the decline in ChatGPT traffic, directly exacerbating negative sentiments towards AI stocks, leading to a wave of selling! Recently, Goldman Sachs analyst Peter Oppenheimer released a report.

A chart in the report shows that ChatGPT traffic has been declining in recent months.

Subsequently, the "Financial Times" reprinted this chart. Now, with the highly sensitive "AI Bubble" story in the financial market, there is new material. Negative sentiments towards AI stocks have once again been ignited!

However, this time, it turns out that the financial industry was too jumpy. Upon closer inspection of this Goldman Sachs analyst's report, there is a major bug. Upon careful examination, the analyst cited data from Similarweb but did not consider the recent change of OpenAI's domain from "chat.openai.com" to "chatgpt.com".

With such a domain change, the traffic of "chat.openai.com" naturally decreased.

In fact, if you look at Similarweb's own data, you will find that it is completely different.

Similarweb's monthly AI services display that ChatGPT achieved a 66.2% year-on-year growth, still the most popular generative AI application to date.

Competitors like Claude and Perplexity are still struggling to catch up but have not surpassed ChatGPT yet.

OpenAI recently announced that the company's weekly active users have reached 200 million, double that of November 2022Not only that, the use of OpenAI models through APIs is also increasing.

Microsoft applications using OpenAI models, such as Github Copilot, are also showing strong growth trends. It can be said that when we take stock of the challenges facing OpenAI, the departure of security researchers and high costs are definitely on the list. But to say that there is insufficient demand is definitely not in line with the facts.

A recent analysis indicates that OpenAI's revenue this year may reach between $3.5 billion and $4.5 billion. For a company that only started serious commercial operations at the end of 2022, this data is impressive. However, the cost behind this is as high as $8.5 billion.

The sequel to "Too much input, too little output"

Recently, Goldman Sachs also emphasized: We are not against generative AI!

In June, Goldman Sachs released a report titled "Generative AI: Too much input, too little output?".

This may have caused misunderstandings, making people think that Goldman Sachs is pessimistic about AI.

In fact, they are just posing a question - note the question mark at the end of the sentence.

To clarify, recently Goldman Sachs analyst Peter Oppenheimer and his team released a sequel, providing a more in-depth explanation of what they understand as "the rational prosperity of technology". In the report, they analyzed it as follows -

Since 2010, the technology industry has generated a global equity return of 32% and a U.S. stock market return of 40%. This reflects stronger fundamentals rather than irrational prosperity.

Earnings per share in the global technology industry have increased by about 400%, while all other industries combined, even including the peak before the financial crisis, have only achieved about 25% growth.

From this perspective, Goldman Sachs does not consider AI to be a bubble, but there is still a hint of warning in itAny good story comes with other risks.

It may exaggerate interest to the point of monopolizing investors' attention, sacrificing other opportunities, leading to unrealistic expectations of future profits, and easily causing a sharp downgrade of the company.

Disruptive technologies almost always go through a process of prosperity - recession - prosperity again.

After the initial bubble bursts, opportunistic followers are eliminated, and only then will large companies do truly useful things.

Goldman Sachs says that every cycle presents the same risks and opportunities -

Historically, investors tend to focus too much on innovators; they underestimate new companies that use others' capital to enter the industry and develop new products and services, as well as the opportunities for non-tech industries to generate higher returns using technology.

Survivors plow the land on the bones of the dead. In Goldman Sachs' view, the initial investment wave and capital expenditure have laid the groundwork for new products and services.

The AI bubble is not entirely the same as in other fields.

Because most dominant companies emerged as "winners" from previous bubbles.

These tech giants profit from advertising, which means there will be a continuous influx of capital expenditure.

Goldman Sachs believes that despite massive investments, there is still significant uncertainty in the development of the AI field. That is, other new players also have the opportunity to carve out a place in AI development.

The "moat" of current winners in the AI field is quite significant, and valuations are not like bubbles, but the rapid growth in the number of new patents in this field indicates that new competitors will emerge, and costs will decrease.

Large enterprises have the ability to invest heavily in AI models, but lower-cost open-source alternatives are rapidly emerging. The world's largest open-source platform, HuggingFace, has about 650,000 models. This indicates that growth in large-scale investments and competitive models is emerging in the AI field, as in previous waves of technological advancement.

Goldman Sachs adds that just as competition is often underestimated, the return on innovative capital expenditure is usually overestimated because over time, the marginal cost of technology will decrease and capacity will increase.

For example, in the early stages of the internet, telecom companies were considered winners.

Compared to more speculative and unprofitable internet companies, they seemed to be a "safer" path to wealth.

But the ultimate winners are those companies that can "ride along" to leverage spending and capacity to build business models.

Many such companies did not emerge until the advent of smartphones and applications in 2006, which then gave rise to growing platform companies, shared mobility, and social media.

In short, generative AI is not a bubbleHowever, in the eyes of Goldman Sachs, today's AI winners will no longer be in asset-light industries.

AI is driving a major capital expenditure boom, and the lack of high returns over the past 15 years, along with current valuations, suggests that this situation will continue.

Moreover, there is currently no evidence to suggest that pioneers like ChatGPT have the ability to sustain commercial success.

In addition, technology is inherently deflationary.

This may help avoid antitrust scrutiny, but it is not conducive to maintaining profit margins.

Therefore, Goldman Sachs advises investors to diversify their investments and not blindly buy more NVIDIA stocks.

Author: Xin Zhiyuan, Source: Xin Zhiyuan, Original Title: "Sora Stuck in Research Quagmire? OpenAI Bleeds $5 Billion in a Year, Goldman Sachs Report Mistakenly Triggers AI Stock Earthquake!"