Organizing | Wang Qilong
In the past six months, hot news such as AI writing novels, painting, and cutting videos have become popular around the world. Now, just by typing a few keywords on the keyboard, AI can draw beautiful pictures while burning our graphics cards. , a whole new world of applications opens up to future startups.
Human beings now have a large number of high-quality, cheap, and fast AI models to generate text, images, videos, software codes, music, sounds, and 3D modeling, and these AI models are free of copyright and plagiarism issues. This opportunity may cause the biggest change in the Internet since the advent of cryptocurrency.
In the past two years, James Currier, a founding partner of NFX, a star early-stage investment institution in Silicon Valley, has invested in 4 companies in this field and plans to make more investments in the next year. This article is based on the article published by James Currier on the NFX blog, and introduces this new revolution in human-machine collaboration to all eager entrepreneurs and developers interested in the potential of AI.
AI’s winning painting. Do you like artificial intelligence paintings?
Names like “Generative AI” or “AIGC” (AI-Generated Content) have become widely known, but in fact, AI only forms half of the equation for this technology. AI models are just the enabling basic layer of these stacks, and the top layer will be thousands or even tens of thousands of applications, an entire mature ecosystem. This article will use a more macro concept - "Generative Tech" (Generative Tech) to name this change.
On January 11, 2021, the artificial intelligence startup Jasper was established. The company uses AI to automatically generate text content on platforms such as blog posts, social media posts, and web pages. Eighteen months after going public, Jasper has revenue of nearly $100 million and a valuation of $1.5 billion.
Today, many "generative AI platforms" have been born around the world, and the generative technology industry is developing so fast that it has been verified by actual revenue and high valuations. Open AI, which supports GPT-3 and other AI models, is also rumored to be raising funding and is valued at billions of dollars. Anthropic, another large AI model manufacturer, has also raised more than US$700 million.
In September 2022, the investment website Signal added "generative technology" as a new investment category, calling on investors or angel investors to join this new field, and encouraging brave founders to create companies in this field immediately to catch up with technology Adopt the best part of the cycle. The new wave has arrived, all it takes is fearless speculators.
What are the unique innovations of “generative technology”?
Source: NFX
Generative technology is a new Internet topology.
To this day, the characteristics of the Internet are still the same as before: that is, the stored old content is first obtained through database query, and then transmitted from the center to individuals at the edge of the information network. Generative technology will change the existing topology of the Internet , because it will first generate unique content at the edge of the information network, and then complete the final realization through individual behavior.
This is a major shift that could bring endless opportunities for startups. If Web1 is "read-only", Web2 is "read-write", and generative technology is "read-write-generate", then Web3 is "read-write-generate-own". Generative technologies are now happening in parallel with Web3 and are developing very rapidly.
Source: NFX
Human activities are transformed by generative technologies.
The existence of generative technology can enable nearly two billion intellectuals around the world to improve their work efficiency and quality. This allows some people to do jobs they could not do before and create new types of jobs. While some jobs will certainly be downgraded, threatened, and replaced—causing fear and self-doubt in tens of millions of workers over the next 36 months—the ultimate outcome will be that the benefits outweigh the disadvantages, and that the overall capabilities of humanity, The gains in productivity and efficiency will far outweigh the losses, ultimately generating trillions of dollars in value.
For intellectuals and creatives, inspiration is crucial. Writers, students, marketers, programmers, architects, graphic designers, musicians, videographers, salespeople, customer service, playwrights... There are too many types of professional workers in the world who are hungry for inspiration, and generative technology Providing a steady stream of inspiration , this tool helps them generate "first ideas" in their minds and achieve breakthroughs from scratch.
Image generated by StabilityAI
In today's era, software has always been used to perfect human beings' initial inspiration, responsible for the production and realization of inspiration after it is born, and turn the fantasies in human brains into reality. Generative technology can also get the help of machines in the process of inspiration formation, combined with the software of the old era, to complete the rest of the way.
This change from 0 to 1 cannot be achieved by today's software. But generative technology allows humans to get AI help from the starting point of any project.
It is precisely because we always encounter creative moments and the birth of inspiration that makes people feel that human beings are so unique and mysterious, and that we are proud of being human. If robots can have thoughts, what else do we humans need to do? This naturally causes a lot of uneasiness.
Of course, like most human-computer interfaces in their inception, humans will eventually get used to it against all odds. In the future, AI will transform from a tool to a human partner, which will be the new normal.
In 1980, Steve Jobs said that Apple's PC was the bicycle of the human mind. Generative technology, then, is a rocket ship for the human mind, and the creators of these AI models may even consider them actual minds.
Humanity has been discussing the inevitability of software-based thinking since the fifties. In 1997, IBM's Deep Blue defeated Kasparov in a chess match. In 2016, AlphaGo achieved a major leap by defeating Lee Sedol in the Go game. Starting in 2022, the advent of generative technologies will have a profound impact on billions of future workers around the world.
Why has generative technology developed so far?
Recently, Open AI has provided an open source alternative specifically for "generative AI models", making generative technology gradually become open source and open in the past six months. In February 2022, Eleuther.ai released GPT-NeoX-20B, becoming OpenAI's open source alternative to GPT-3 generative text technology; half a year later, in August 2022, StabilityAI launched Stable Diffusion, becoming OpenAI DALL-E 2's open-source alternative to generative image/video, and not long ago, Stable Diffusion announced $100 million in financing.
Generative technologies have gradually attracted multiple companies competing on price, quality, and convenience. As a feedback, the cost of AI-generated images has dropped 100-fold in the past two months. Today, the trained Stable Diffusion model can fine-tune and imitate Hayao Miyazaki's famous Ghibli style to the point where it looks like the real thing.
StabilityAI has trained a mature model in the animation style of Hayao Miyazaki
It can be seen that the quality of texts, images, codes, and voices generated by AI is rapidly reaching human levels. Some of these works can even make people mistakenly think that AI has passed the Turing test. The quality of works of art is usually difficult to quantitatively compare or objectively measure, but when you see a work of art, you will subconsciously react to it immediately and make a quality judgment.
As the 2021 Stanford Artificial Intelligence Index points out, for some limited applications, AI has developed to "a high enough standard that it is difficult for humans to distinguish between AI-synthesized and non-AI-synthesized works." We are moving in the direction of "everything can be generated".
This is a Cambrian explosion in technology. The ecology of the Internet has begun to change with the birth of generative technology, and the information industry wave will come again. Generative technology is still in its early stages, but it’s already evolving from the stuff of science fiction to reality. Generative technology is different from self-driving cars because it does not face regulatory issues, let alone worry about traffic accidents and life safety. Generative technology is different from virtual reality, because virtual reality technology still needs to find meaningful application methods, and it does not need to popularize hardware devices. Generative technology now has a very mature application method-that is, creating inspiration.
In fact, there is no need to consider cost and quality. Almost any kind of generative technology content is enough for a company to develop: Podcast.ai is a podcast that operates entirely through "generative technology", October 12, 2022. In the first podcast episode launched by Podcast.ai, AI simulated the late Steve Jobs, allowing the well-known American podcast host Joe Rogan to have a 20-minute conversation with the "AI Steve Jobs".
Generative technology is still just a slope, and it still needs to climb relentlessly to achieve changes in the technology world. After 14 years of stagnation in AI technology, most of our software and human-machine interfaces will likely start to be significantly enhanced within the next five years, providing endless opportunities for entrepreneurs.
What are the characteristics of generative technology products?
Generative technology products are divided into two layers:
The bottom layer is naturally its AI model. AI models can absorb users' unique ideas and generate novel results, such as OpenAI's DALL-E or GPT-3, which are typical AI models. The price of AI models is still gradually declining. In the early days, it was still unattainable at US$100 million, but now it has dropped to the tens of millions. And open source AI models are springing up like mushrooms after a rain, and they will surely be within reach in the future.
The top layer is its application. Applications are the result of everything that generates lasting business and creates value for the company. If startups want to gain a competitive advantage, they not only need unique ideas, but also qualified packaging and applications, both of which are indispensable.
Source: NFX
Generative technology lowers technical barriers in many fields. Even if a person is not an architect, he can know what kind of house is suitable for him and use this technology to generate house renovation drawings; even if a person is not an illustrator, he can know what kind of paintings are beautiful and use this technology. Technology generates AI paintings... On this basis, generative technology companies can be summarized into three core characteristics:
1. From 0 to 1 - from 0 to 10!
As mentioned earlier, generative technology can solve the problem of lack of inspiration that humans have struggled with for centuries. AI can help countless creative professions obtain inspiration and achieve breakthroughs from 0 to 1; however, from 0 to 1 only AI can do The first thing is that after providing inspiration, you will find that even the remaining work can be completed by AI. AI will be able to provide solutions from 0 to 10 - in other words, as long as you have the corresponding AI model, the product will be able to meet all the needs of users.
Since AI can generate inspiration, why not just let AI do the rest of the work?
In fact, such a generative technology project has already appeared, that is, SALT. SALT is the world's first AI-created movie, entirely written and dubbed by AI. The AI is modeled after 1970s science fiction movies, using a combination of generative technology tools to quickly generate the final video (Stable Diffusion, MidJourney, DALL- E), writing scripts (GPT-3) and generating character voices (Murf, Synthesia).
2. Replace curation with creation.
The so-called curation means planning, selecting and displaying. In the two decades since the turn of the millennium, humanity has been pursuing personalization through curation; e-commerce vendors, video sites, and music sites of all kinds serve you from their central databases day after day. Selected products that you are most likely to like, which are so-called "personalized recommendations" and "big data push".
Curation is a very limited approach to personalization that is based on calling on existing data and trying to incorporate people’s preferences into existing products, rather than creating new products that are best for users.
Now, generative technology is changing that, and it’s personalizing in ways we’ve never experienced before. Generative technology replaces curation with creation, does not require complex database calls, but is trained by the database and generates completely new things.
Are these all empty words? In fact, generative technology currently has a definite impact on the development of some fields, including the field of music. There are now many platforms - Boomy, Amper, Aiva, Ecrett, SoundDraw - that allow people who have never made music to create original songs in seconds. The Boomy platform also provides Creators provide tools to share and monetize these works, allowing songs to be released directly on Spotify, TikTok, YouTube and more than 40 other platforms around the world.
AI is gradually changing our lives. In the past, we might choose a popular playlist to play while traveling by car, or blindly follow "personalized recommendations" and "big data push" to listen to music; but in the not-too-distant future, in the world of generative technology , we will be able to generate and listen to the songs we really need anytime and anywhere according to our occasion, current mood, blood pressure, and heart rate.
3. Harmless and low-friction interface.
The biggest feature of generative technology tools is their ease of use. Such heavy work is all done by AI models, eliminating friction to the greatest extent during the creation process. Mainstream AI models like DALL-E and Stable Diffusion can generate jaw-dropping works of art in 30 seconds with only simple text prompts.
In the future, generative technology will become routine and even automated. Imagine that one day in the future, two people walk into a house in a virtual community. Generative technology can continuously generate new things based on the personalities of the two users-these things can be artwork, games, music, and shared experiences. topic, and if these two users plan to trade, the generative technology can also automatically generate NFT for them to choose...
This will be another revolution in human-machine collaboration. AI has transformed from a tool into a human partner, inspiring people while bringing endless surprises.
What do generative technology companies do?
Now that everything is handed over to AI, what do generative technology companies need to do? This is an easy misunderstanding to fall into, because every great idea using generative technology ultimately comes from the people who control the AI model. Think about the possibility, what would happen if one or more basic layer AI models were combined?
Source: NFX
Innovation is quite difficult. Every year, there are thousands of articles teaching entrepreneurs how to innovate, how to find their own advantages, and how to seize opportunities... However, using AI to innovate is quite simple . Because entrepreneurs can directly look for an existing field that has not yet used AI on a large scale, directly leading their peers who are still in the handicraft industry era (this sentence is a bit exaggerated - however, for future AI, humans may It’s really still in the handicraft era).
Seizing the opportunity can create an advantage, but this advantage will quickly dissipate with the entry of large companies, the emergence of more general AI models, and business competition. Now comes the topic. What role do generative technology companies need to play at this time?
Think about how each new user can add value to every other user.
AI models can be used by anyone. Just type in keywords, and you and I can use AI to generate beautiful pictures. Therefore, in addition to the technology itself, the most important thing for generative technology companies is the community. The community can create value and create more network effects.
Think about how generative technology can be integrated into your business or personal life.
So far, generative technology is still in the "picture-and-leaf" stage, and not many people have actually used it to create and let it help make their lives better. Generative technology companies must play a guiding role so that users do not want to leave the help of AI models in the long run.
Combined with inventions from the old days.
AI models won't really replace all existing software, we still rely on browsers and applications every day - so why not just combine these things with new-age AI models? It is impossible for people to rely on an AI model to "make a difference", and it is impossible for the public to accept the changes of the new era instantly. Therefore, the best way is to integrate the AI model into the software we use every day, so as to exert a subtle influence on it. role.
How should entrepreneurs seize this opportunity?
Technology and market are like a river. Some parts of the river run faster, some parts slower, and there are eddies in the river that can pull you backwards while being sucked in. The trends that emerge on this river are created by technology, market segments, language, distribution channels, teams and beliefs.
As an entrepreneur, if you want to stay on top of this wave, you must act now—not in the next six months or the next three years, but this month or even this week. Because the river flows so fast that people have to pause what they are doing and focus on thinking about its problems.
In the past two years, James Currier, a founding partner of NFX, a star early-stage investment institution in Silicon Valley, has invested in 4 companies in this field and plans to make more investments in the next year. Next, we will take the four companies Currier invested in as examples to see how he applies generative technology:
Gaming company Latitude.
At the beginning of 2021, NFX invested in the gaming company Latitude. This startup is committed to developing the "AI Dungeon" (AI Dungeon) game generated by artificial intelligence, and announced at the beginning of the project that it had raised $3.3 million in seed funding.
Generative technology will bring a new gaming experience, which is different from the classic text adventure games we used to play. In previous games, the dialogue of the characters was completely arranged by the game designer, and the player could only choose from a few options - and " The characters in "AI Dungeon" can respond to any command of the player.
Generative technology can not only add dialogue to NPCs in old games and make up for the plot shortcomings of game works for fan creation, but it can also give birth to new types of games like "AI Dungeon". Latitude is leading the way in AI gaming with its newly released Voyage gaming platform. The startup has been working in the core area of generative technology for 3 years and plans to work with AI21 Lab, which researches generative writing technology, to further integrate natural language processing technology into their platform.
AI Dungeon game interface, enter text in the box below to talk to the game character
Legal claims search service Darrow.
Every day, millions of people around the world are unknowingly harmed by illegal activities—air pollution that causes cancer, privacy breaches that jeopardize sensitive information, the occasional unsafe snack or meal, car accidents, Vehicle accidents caused by defective parts, or even defective drugs that caused irreversible damage... Once upon a time, it was almost impossible to find these irregularities in massive amounts of data, let alone build a case around them. However, the birth of AI It has completely changed the difficulty of data search in this era.
Darrow scans publicly available real data from all corners of the web, including official sources from social networks and news media, as well as sources of litigation and financial information. Darrow’s AI can then connect all relevant data points within the context of this information to detect violations. Finally, Darrow builds a strong case for detected violations and addresses the hidden dangers discussed above.
The.com, a platform for automatically programmatically generating websites and pages.
The name The.com sounds like a very rare domain name, and this website is as unique as the name. It does not engage in AI art creation, but uses websites to generate websites. In short, The.com is a generative platform that can code and develop hundreds of websites with simple database calls and lightweight prompts. However, anyone can do website development. The real killer feature of The.com is that it can integrate other websites, add thousands of pages to your own website, and expand the traffic of the website.
The.com's goal is to modernize website building while ensuring that web creators receive credit for their work. The company, now armed with $4.4 million in seed funding from NFX, has created what it calls a "low-code website building platform" that aims to move away from the template-based approach that has become the industry standard. Simple modularity and open source are naturally things we all like to see. The.com users can even collaborate with each other and chat directly during the website construction process to achieve common traffic.
Real estate technology website Tailorbird.
Every property in the world is unique, so a lot of time is often wasted in the work of multifamily renovation between managing stakeholders, budgets, organization and implementation; since the beginning of 2020, the AI and IoT craze has swept the world , infrastructure is developing rapidly, and a new industry has begun to emerge in the United States, which has caused great changes in the real estate landscape in the United States. This industry is real estate technology (proptech) .
Real estate technology refers to companies that use innovative operations and technology to create efficiency and value in the buying and selling, leasing, marketing, operations management and finance of various space asset classes. Tailorbird uses deep learning to build new floor plans for homeowners looking to renovate.
These floor plans are generated from photos and floor plans taken on the web, and AI collects this data and combines it with remote 3D measurements to generate many new renovation plans. These AI-generated plans quickly provide designers and homeowners with great inspiration and ideas, and reduce the time and costs associated with home renovations.
Look, generative technology is not just about using AI to draw pictures and write some essays. This field is still in its infancy and has unlimited room for development. Entrepreneurs often worry about how to make their products unique and make a splash in the market, but in today's era, just adding generative technology to existing products can already make them unique.
How is this article useful to me?
Looking back at the history of the world, major technological changes always unfold slowly: until 2003, many people in the world were still skeptical of the Internet, so the first pioneers who established Silicon Valley were able to avoid fierce business competition; from 1997 to In 2005, it took nearly 10 years for SaaS (Software as a Service) to gain industry consensus; Apple only opened it to external developers 18 months after releasing the iOS platform... Web3 has been launched 10 years ago.
But generative techniques are different. Almost everyone who sees the news about AI creation will realize one thing: when the technology matures, it can directly change the existing creative industry. Countless venture capital companies, startups, entrepreneurs, and commercial companies have discovered the fact that as long as they continue to "feed" AI models, generative technology may change the world tomorrow.
All in all, time is tight for entrepreneurs. If you want to take a share of the rapidly developing field of generative technology, you must act now; entrepreneurs now need to decide who their target customers are, where their distribution channels are, and what ideas they want to create with AI... Today, there are Many fields have not been involved in the wave of the AI era, and there are many choices.
What do great founders of the generative technology industry look like?
They look like you. Generative technology has subverted the history of human creation. In the era of AI, anyone with an idea can start a business.
Reference article: https://www.nfx.com/post/generative-tech
☞微信新增图片、视频拖动一键发送功能；Google 称互联网的六成是重复内容；Angular 15 发布|极客头条
☞Meta发布 “科研者的福音”，上线仅三天被骂到撤退
☞华为开发者大会 2022 开发者主题演讲与技术分论坛干货分享（附课件）