The explosion of ChatGPT has ignited the expectation of generative AI in the technology circle, but for players in this field, opportunities are accompanied by a huge impact. Among them, the most influential one is Jasper.ai, a cutting-edge AI unicorn.
Before ChatGPT, Jasper was the most promising AI text generation tool. It can create hundreds of words of text based on a simple phrase or prompt, making it popular among media workers and marketers.
In January 2022, the Jasper team had only 9 people, and it expanded to more than 160 people 10 months later, with annual revenue expected to reach $60 million. The capital market welcomes Jasper at a glance. In October 2022, Jasper raised US$125 million in financing and reached a valuation of US$1.5 billion just 18 months after its establishment. Thomas Laffont, co-founder and senior managing director of Coatue Capital, described seeing the Jasper demo for the first time as a “magical experience”—like seeing the first iPhone, which convinced him that the world was about to Huge changes.
However, only 1 month later, ChatGPT came. “No one thought that the free ChatGPT could be so good and so fast.” The high-spirited Jasper thought that the final victory was at his fingertips, but in just a few months, he experienced a transformation from heaven to hell.
To make matters worse, there is a subtle relationship between ChatGPT and Jasper. ChatGPT is a chat robot application based on OpenAI’s own GPT-3 large model; and the underlying technology of Jasper is also OpenAI’s GPT-3. In other words, Jasper built its own business on the opponent’s platform. Right now, Jasper is inseparable from OpenAI, but has to face the dilemma of “platform grabbing meat”.
Based on this background, the technology media The Information recently went to the Jasper.ai office to visit CEO Dave Rogenmoser. They reviewed Jasper’s entrepreneurial story and evolution process, and asked Rogenmoser to tell them about the impact of ChatGPT in the past month. Some readers believe that this is one of The information’s best reports in recent years.
The following is the full text of the report, the original title “The Best Little Unicorn in Texas: Jasper Was Winning the AI Race—Then ChatGPT Blew Up the Whole Game”, the original author Arielle Pardes. Compiled and organized by Geek Park.
01 “We don’t try to compete with our partners”
A few weeks ago, Sam Altman, CEO of artificial intelligence startup OpenAI, made a Zoom call with Dave Rogenmoser, the CEO of Jasper, on the other side of the screen. Jasper is a copywriting startup based on OpenAI’s flagship large-scale language model, GPT-3. The two companies have a shared Slack channel where they exchange updates and feedback on GPT-3, the backbone on which Jasper’s business depends.
OpenAI is an artificial intelligence research laboratory comprised of OpenAI LP , a for- profit company, and OpenAI Inc, a not-for-profit parent company. ｜Image Source: Visual China
Rogenmoser scheduled the call with Altman because he had just seen a post on the Slack channel where OpenAI announced ChatGPT—a new product from OpenAI that works almost like magic. With just one simple prompt, ChatGPT can draft a business proposal, write a resignation letter, or explain the inner workings of quantum mechanics. In fact, it works a lot like Jasper, the company’s core product. But unlike Jasper, ChatGPT is free.
Rogenmoser has some questions about this. “Look, we need to know something about what you guys are going to do,” he told Altman at his office in Austin, Texas, where Jasper, an OpenAI partner who paid handsomely to use GPT-3, told Altman. OpenAI doesn’t owe Rogenmoser anything, but if it plans to give the chatbot away for free, the heads of Jasper will need to rethink things.
According to Rogenmoser, Altman assured him that ChatGPT would not be free forever. ChatGPT became an instant hit, amassing over 1 million users within a few days. Ultimately, OpenAI will have to put it behind a “paywall,” if only to pay for the staggering computational costs. Each chat query costs OpenAI a few cents, a fee that adds up quickly when a million users use it multiple times a day.
Altman also told Rogenmoser that he didn’t expect ChatGPT to be so popular. He sees it more as an interface upgrade than a technological revolution, and marvels at its explosive popularity. According to Rogenmoser, Altman offered a bit of reassurance: “We’re not trying to compete with our partners.
Jasper is used to automatically generate copywriting
Still, Rogenmoser has a sense of urgency. In just a few months, Jasper has become a rising star in the exploding field of generative AI.
In October 2022, Jasper raised $125 million and reached a $1.5 billion valuation just 18 months after its founding. Its customers have grown to 100,000, three-quarters of whom pay $80 or more a month for Jasper’s suite of advanced AI-powered writing templates. According to The Information, the company’s 2021 revenue is $30 million and is expected to double in 2022.
In contrast, OpenAI’s revenue is only tens of millions of dollars–its revenue mainly comes from licensing fees for startups like Jasper to call its API. According to Reuters, OpenAI expects these partnerships to generate $200 million in revenue by 2023. An OpenAI spokesman declined to comment directly on the company’s revenue.
Before ChatGPT came along, Jasper had unconditional praise from customers: One user said it did basic writing tasks even better than entry-level employees, proving she was a monthly contributor to Jasper’s “boss mode.” The $100 paid for the product is reasonable. The product offers long-form writing templates and the benefit of generating up to 50,000 words. Another client, a web developer, said he had moved away from hiring freelancers to write web copy and started using Jasper instead. “The potential is limitless,” he added.
Jasper’s growth matches this potential through November 2022. In January 2022, the company had only nine employees. By November, it had more than 160 employees, including more than 70 new engineers and designers tasked with building more and better custom templates.
This article, “What’s The Big Deal with AI Right Now”, was written by Jasper.ai with the following prompt: “Write an article about AI and how it is now generating content for the first time ever.” | Image Source: Coatue
Most of Jasper’s current 60 templates are tailored to the needs of marketers who use the service to generate blog posts, press releases, social media headlines and ad copy. There are also templates for brainstorming product names, generating viral tweets, or distilling long articles into short bullet points. In addition, for those users who need more divergent content, there is a free mode option, which allows Jasper to write anything from scratch.
The company is also developing features to make Jasper accessible outside of its own platform. Like a Chrome extension that brings AI-customized copies to other parts of the web. “We want to enhance every input box on the Internet,” Rogenmoser said in November. “Not even on the internet. Every word you write, I’m there to help.”
The Shock of ChatGPT
However, ChatGPT appeared. Within days, it was revealed that this was not just an advance in generative AI, but possibly a black swan event. “No one thought ChatGPT would be this good, this fast, and it would be completely free,” says Chris Frantz, co-founder of content marketing startup Snazzy AI.
When products like Jasper and Snazzy launch in early 2021, GPT-3 will still need high-precision cues to give good results. This makes copywriting startups all the more important as they bridge the gap between the complexity of AI language models and the utilitarian needs of end users.
But ChatGPT has completely changed this relationship. “You can say to ChatGPT, generate a blog post about X, and you get good results. That’s all you have to do,” Frantz said. “As a result, niches based on prompt demand (like Jasper and Snazzy) suddenly disappeared.” Not to mention, who pays $80 a month for Jasper when users can get ChatGPT for free?
That moment, Rogenmoser said, “really got on everyone’s nerves.” Within days, he had assembled a team to come up with a response plan that included Jasper building out his own chat feature — one that, like ChatGPT, could respond to questions conversationally. (Note: This feature launched on December 20th)
At present, Jasper has responded quickly and launched more original functions, such as text generation pictures, chat robots, etc. ｜Image source: https://www.jasper.ai/
The team started thinking about building more enterprise capabilities for existing clients at big companies like Airbnb, HarperCollins, and IBM. Jasper also completed the acquisition of Outread, a browser extension that improves the grammar and fluency of applications such as Google Docs and Microsoft Word. Here’s another bonus: Outright already has 1 million users, 10 times the user base of Jasper.
In just a few days, Rogenmoser has gone from a seed player in the text-based AI startup space to an embattled player needing to defend his turf. Jasper caught the generative AI wave early and has so far proven itself capable of riding it. But now the company has to prove it can stick around for the long haul.
02 Jasper’s Evolution and Controversy
Two days before ChatGPT’s surprise launch on November 30, Arielle Pardes, author of The information, visited Rogenmoser at Jasper’s office in the western suburbs of Austin.
From the outside, the building looks like an ordinary office park. But inside, the Jasper headquarters reminds her of the startup offices that began popping up in San Francisco’s SoMA district a decade ago. There are foosball and ping pong tables, snacks in glass jars and cafeteria-style tables where employees meet for lunch each day. A conference room was converted into a golf simulation practice room. Rogenmoser refuses to give himself a dedicated space, instead adopting Zuckerbergian style, sitting at a desk in the middle of the open office.
Rogenmoser is very tall, 6 feet 8 inches (2.07 meters), but his towering height contrasts with the excitement of the kids who ran to homecoming as a child. For a CEO of a billion-dollar company, he’s remarkably candid. When asked about his experience studying marketing at Kansas State University, Rogenmoser admitted that he was more focused on being the best fraternity guy on campus than on his academics. He said he skipped many classes and occasionally cheated on exams.
Rogenmoser described, “Jasper is radical!” This is different from Altman. Altman founded OpenAI out of a deep belief that AI will open an era of “abundance” that will benefit all mankind. Rogenmoser just wants to Get rich — or, as he puts it, achieve “a measure of financial freedom.” He looks back on the rough eight years leading up to Jasper’s breakthrough as a story of perseverance.
His original entrepreneurial idea was to make a “fraternity” recruitment software, which was related to his college experience. But by his own admission, it was a bad idea. He later started a small marketing firm with two college friends, JP Morgan and Chris Hull, providing copywriting services for website content, Facebook ads and search engine optimization. Neither Rogenmoser nor his friends have these skills, so they outsource the work to contractors and take a cut of the profits.
Morgan, Rogenmoser, and Hull spent eight years trying out startup ideas and eventually made their fortunes together with Jasper. ｜Image source: The Information
Morgan and Hull eventually became co-founders of Jasper. Morgan, the only one of the trio with technical skills who taught himself to program, built Jasper’s back end and serves as CTO. As COO, Hull is responsible for people and operations. The three knew from each other that they wanted to work together, shared a love of golf, and shared many Midwestern values—they have 10 children together.
After starting a marketing agency, they created an online course to teach others about marketing. Through the course, they gained access to the email addresses of thousands of people working in advertising, communications, and sales. To help market their courses, Morgan built a small pop-up window on their website to increase conversions. Because it worked so well, they turned it into their own business called Proof.
In 2018, Proof was accepted by Y Combinator, then raised $2 million and grown to 15 employees. But the startup has struggled to scale. “We never knew how to go from a feature to a platform,” said Rogenmoser, who fired half of Proof’s staff in late 2020 and promised to rehire them when the company figured out how to fix the business.
That’s when Rogenmoser saw people talking about GPT-3 on Twitter. OpenAI introduced the model in the summer of 2020, and AI insiders were almost salivating at what it represented. Built on the Transformer model, GPT-3 is a neural network that can infer relationships between large amounts of data scraped from the internet.
Transformer models offer “a broader capability around natural language communication,” says Michael Wooldridge, who teaches computer science at Oxford and is the author of “A Brief History of Artificial Intelligence.” OpenAI has built two models before this, but GPT-3 is known for its huge training set. “It trained on 500 billion words — basically the entire World Wide Web,” Wooldridge said.
In late 2020, OpenAI opened a small private beta to developers to see how they could use the technology in the real world. Rogenmoser used his connections at Y Combinator (Altman was a former Y Combinator partner, though he and Rogenmoser never met) to gain access in December, although the beta slots were very limited. Only a handful of development teams have access to GPT-3, giving Rogenmoser and his partners a significant head start on the product.
Once they started experimenting with GPT-3, they had a clear idea of what they wanted to do with it: reach the same customer base as they did with Proof, but give them more functionality. Morgan made a prototype, which the team dubbed “Conversation AI,” and Rogenmoser took it to some enterprise customers to see what they thought. “I’ve been selling this personalization software for two years, and it’s just a bug,” he said. “But with this, people are falling off their chairs.”
“Conversational AI” was launched in January 2021, and the team later changed the product name to Jarvis, named after Tony Stark’s AI butler in “Iron Man”, but after receiving After Marvel’s order to stop notifications, it was renamed Jasper.
The cost to use the platform is $29 per month, and there are 8 templates at launch. In fact, it is not the first company to apply GPT-3 to the market. Other startups, such as Copy AI, have already begun to use GPT-3 to write copywriting. But based on the accumulation during the Proof period, Rogenmoser has a huge address book of marketing professionals that can be developed as potential customers.
Selling Jasper is as easy as showing people how to use it. Thomas Laffont, co-founder and senior managing director of Coatue Capital, described his initial demo of Jasper as a “magical experience” that, like seeing the first-generation iPhone, convinced him that the world was about to change dramatically. Coatue has invested in other AI companies before this, but Laffont likes Jasper’s team, which he says is “competitive and broad-minded.”
In 2018, the three founders “spent a lot of time in luxury bathtubs dreaming of conquering the world” during their Y Combinator demo day. ｜Image source: The Information
In October of the same year, Coatue participated in an $85 million Series A round of financing, along with Insight Partners, Bessemer Venture Partners and others. That makes Jasper one of the fastest startups in history to reach unicorn status, and it’s an extraordinary case in a year that’s been sluggish for funding.
“They got a huge opportunity,” Laffont said. “Jasper is in a leading position with a clear product-market fit. They’ve also landed a great set of investors with a variety of expertise.” Even more, they have thousands of customers eager to harness the magic of generative AI. All signs point in the right direction. Then, out of nowhere, OpenAI entered the world of chat apps.
For half a century, futurists have imagined a world in which technological advances lead to the complete fusion of humans and machines. In this world, intelligent machines might effectively evolve Homo sapiens in the same way that the use of fire led to the leap in human intelligence. Computer scientist JCR Licklider described this phenomenon as “human-machine symbiosis” in a famous 1960 paper. Rather than computers replacing humans, Licklider envisions their relationship as a partnership where “humans will set goals, formulate hypotheses, set standards, and evaluate. Computers will do programmable work.”
We have never seemed closer to this reality. The development of GPT-3 has opened up a dizzying array of possibilities: it can instantly generate hundreds of lines of code, the perfect Shakespearean sonnet, or make suggestions for resolving geopolitical conflicts; it can do better than SparkNotes Write news articles, draft academic papers, and refine fiction themes efficiently.
The impact of a powerful text generator like GPT-3 is far-reaching, and the onus of protecting its users falls largely on its maker, OpenAI. But the startups building the infrastructure for this new world will also face decisions about who AI will serve and how it will be used.
Early on, Jasper positioned itself as a tool to improve copywriting productivity. Rogenmoser said, “All our marketing is about how we make you successful. In the hero’s journey, our customers are the heroes.” But it didn’t take long for users to come up with less heroic use cases for the product. Some people use Jasper to help with homework or write school papers. “It’s a great way to make sure your essay is original and free of plagiarism, and it can be written on any topic in whatever format you specify in advance,” wrote one blogger.
When asked how Jasper would respond to his product’s potential to “disrupt” education, Rogenmoser seemed surprised by the question. “It would be too bad if we made everyone dumb,” he said.
Products like Jasper don’t herald the end of thinking, he argues. There are other ways of assessing learning, such as oral presentations or Socratic debates. Maybe school papers won’t survive the AI revolution, but Rogenmoser doesn’t think that’s a bad thing, “I’d say people 20 years ago probably thought learning perfect spelling was valuable learning, and if you couldn’t spell correctly Something, and you can’t live a good life.” The advent of spell-checkers challenged that assumption, but they didn’t completely replace thinking or learning.
“I’m definitely not in favor of cheating,” he added. “But things are changing. I think schools have to figure that out.”
As for other “grey” uses, a customer success representative for Jasper said she occasionally sees people trying to use it to generate pornography and other overtly sexual content. In response, Rogenmoser said these use cases violated Jasper’s content policy, which also prohibits defamation, discrimination and illegal content.
The history of emerging technologies suggests that it may only be a matter of time before Jasper introduces fake news, racist conspiracies or misogynist bias into professional marketing campaigns. Rogenmoser seems overly optimistic when it comes to these concerns about how people are using Jasper, as does OpenAI’s Altman. Rogenmoser said he was “unaccustomed” to cynicism, and rattled off several great things that generative AI could promise, such as helping people with dyslexia communicate better.
He paused and said, “I do worry that Jasper will become a company that doesn’t actually do anything good, and Facebook might be in that camp. Once invested in it, it would almost be better if the company didn’t exist … I don’t want to accidentally find myself in this situation a few years later and think—damn, why are we doing this?”
03 Rely on Open AI risks and ways out
If you put Rogenmoser and Altman together, you’ll notice more than a few similarities: They’re both white men in their 30s who grew up in the American Midwest. Both self-proclaimed idealists, both are CEOs of startups and shareholders of Tesla. They all got rich in the AI gold rush.
Sam Altman , CEO of Open AI ｜Image credit: Vox
But while Altman has become a whale in the AI open ocean, Rogenmoser is more like a remora, a finned fish that attaches to cetaceans and feeds on debris. OpenAI needs partners like Jasper to pay the bills, but not nearly as much as Jasper depends on OpenAI.
Joanne Chen, a partner at Foundation Capital, said Jasper’s reliance on OpenAI was one of the “biggest risks” she flagged before investing in its seed round in 2021 and its Series A in October 2022. “But since then,” she said, “I think the risk has been significantly reduced.”
Chen is encouraged by the number of other large language models being developed by companies like Meta, Amazon, Nvidia and Google. Jasper has already begun to integrate other open source models, such as GPT-J, into its products. “These models are getting better and better, and we’re having more options.”
For Jasper’s customers, sticking with the service will largely depend on whether it performs better than competitors. According to Ayrton Mendoza, a designer and web developer at online marketing agency Postweld, Jasper still outperforms ChatGPT on copywriting tasks because of the way Jasper’s templates are designed.
Mendoza has hired freelancers on sites like Fiverr in the past to write marketing copy for the sites he produces, but the process has been expensive and time-consuming. “It saves a lot of time,” he said of Jasper, “you can write an entire article in minutes.” Mendoza’s favorite feature is Jasper’s ability to change the tone, make an email sound cheerful, or make ad copy more humorous.
“I can make these adjustments at any time without having to send back and forth messages like when I hire a freelance writer. I need to constantly ask, can you say this? Or, can you use a different tone?” Mendoza said. “I also tried asking Jasper to write like Joe Rogan, or write like Elon Musk.” The results were interesting, sometimes surprisingly accurate.
But other customers say Jasper’s product works well, but not enough to justify its price.
Grant Tucker, founder of creative studio Chromaspring, said: “I used Jasper for a few months, but I was able to get better content output from ChatGPT in 10 minutes.” Love ChatGPT’s conversational interface, which makes modification easier.
Recently, he asked ChatGPT to summarize a campaign he was working with Adidas on. “It gave me this rough brief that I was able to tweak with instructions — make this longer, don’t talk about how close the relationship is,” he says. Overall, he was pleased with the experience. “ChatGPT has no bells and whistles compared to Jasper, but for the vast majority of us, it’s all we need.”
Rogenmoser thinks that’s great. He sees Jasper’s future as more of an enterprise platform, like Atlassian or Slack, where companies might pay to integrate it into their workflows. “ChatGPT can do a lot of things, but it lacks, and will likely continue to lack, all the features that big businesses want,” he said. “We see ourselves as a company that helps businesses introduce AI.”
Other startups hope to do the same, bringing to market a range of new enterprise products that bring GPT-3 to email, text messages and web browsing. Rogenmoser is optimistic and not too worried about this competition. “We get to that level by being fast and aggressive,” he told me. “That’s how we’ll earn a spot six months from now.”
This article is transferred from: https://www.geekpark.net/news/313662
This site is only for collection, and the copyright belongs to the original author.