生成式AI圣经：生成式AI颠覆的终极指南

The ultimate guidetogenAI disruption Unleash thegenerative AIrevolution Dive into a treasure trove of generative AIcompanies, investors, and market insights.Elevate your decisions with exclusive softwarebuyer interviews to understand product pricing,CSAT, and post-deployment experience. CB Insights–Your Gateway to Generative AI Sign up for access Generative AI is a theme CB Insights covered before thehype.We've consistently been ahead of the curve And we'restaying ahead today–so you can too 2023 Stay on top of the landscape with research & transcripts Understand 25+ generative AI markets Large language model developersCustomer support operationsText generationProtein & drug designVoice synthesis & cloning+ more Discover 400+genAIvendors Trackedin our analyst-curatedExpertCollection 4 ContentsGenerative AI Bible 63So where is generative AI headed? 641. Race to dominategenAIinfrastructure Then suddenly19 882. Cross-industry applications face pressurefrom large players And now suddenly is accelerating37 Hundreds of startups pile intogenAI38 Funding soars as investors flock41 Big tech is all in and ready to fight50 119Promising companies to watch Thegenerative AIboom “How did yougo bankrupt?” “Two ways…Graduallyand thensuddenly” Gradually Generative AI has been in the works for years How did we get here?A recent timeline of selectevents in the development of generative AI GANs tap into the idea of “AI versus AI”—advancingimage generation dramatically Images from 2018 paper where DeepMind researchers trained GANs on a large-scale dataset to create “BigGANs” Imagesource:Large Scale GAN Training for High Fidelity Natural Image Synthesis*”AI versusAI”:A breakthrough where two neural networks try to outsmart each other,creating and refiningsyntheticoutputs. WaveNetproduces synthetic audio, showcasing thepotential of generative models beyond images The Transformer architecture can better understand andgenerate human language, paving the way for further R&D 6 authors of the seminal research paper have gone on to raise $1.7B across 5AI companies* Google AI releases BERT, a leap in the ability of machinesto understand context in language The AI language model predicts a word based on not only the preceding words, butalso the succeeding ones (bidirectional understanding of context). OpenAIincorporates the Transformer architecture into itslanguage models OpenAI’sGPT-3 is a major leap forward, showcasing abilityto generate code, jokes, and more Deepfakes go mainstream, highlighting power and pitfallsof video generation GPT-3 is the foundation for DALL-E, which can generateimages from text descriptions Then suddenly GenAIgoes from experiment to everywhere Models get bigger… …and better, beating human performance benchmarks Text-to-image generators take the internet by storm Generative AI startups raise major funding to fuel growth Deals worth $100M+ to generative AI startups in 2022 AI coding assistant GitHub Copilotbecomes widely available ChatGPTgoes viral, getting to 1M users in 5 days and100M in 2 months—unleashinggenAIboom Time to 1M users for select platforms/apps from launch Almost overnight, exec interest in generative AIskyrockets and companies feel pressured to react Earnings call mentions of “generative AI” (as of 9/30/2023) Microsoft invests $10B intoChatGPT-makerOpenAIinQ1’23, propellinggenAIfunding to new heightsDisclosed equity funding & deals to generative AI companies (as of 9/30/2023) Microsoft debuts AI-powered Bing, running on a newOpenAI large language model (LLM) Meta introduces Llama, an open-source language model Google sounds alarm bells and releases Bard chatbot On the AI side, it is a really exciting time. I think we've beeninvesting for a while, and it's clear that the market isready…Obviously, we need to make sure we're iterating inpublic, these models will keep getting better, so the fieldis fast changing.The serving costs will need to beimproved. So I view it as very, very early days,but we are committedto…actually bringing direct LLM experiences in Search,making APIs available for developers and enterprises andlearn from there and iterate like we've always done. Alphabet CEO Sundar PichaiQ1’23 earnings call GPT-4 becomes OpenAI’s most powerful model yet,crushing human exams Reddit, after providing years of free training data for AIsystems, plans to charge for access to its content Chegg blamesChatGPTfor declining revenue, sees itsshare price tank, and, in response, pivots to build LLMs Nvidia enters the $1T market cap club as demandfor GPUs used in genAI sends its revenue soaring US companies to reach $1T+ market cap (as of 10/30/2023) Stack Overflow sees declining traffic and lays offemployees amid AI coding boom Even Apple scrambles as it works on its own LLM calledAjax and is on course to spend $1B a year ongenAIpush If you take a step back, we view AI and machine learning ascore fundamental technologies that are integral