首页 > 译 文 > 2024年对产品经理的创新有借鉴价值的40+个大想法(下)



Fintech | 金融技术

24)The Rise of the Developer as Buyer in Financial Services


In 2024, the developer will become one of the most important influencers in the purchase of financial services infrastructure.

Historically, financial services infrastructure purchases were mostly driven by the economic buyer (“What’s my ROI?”) or the business lead (“Does this solve my use cases?”). But there is now a third, increasingly influential constituent: the developer. Look no further than Moov’s growing fintech developer conference and associated 4,000+ strong community. In tandem, developers at larger financial institutions and insurance companies are becoming more influential in buying decisions.  

The rise of the developer as buyer in financial services companies of all sizes favors new entrants; for fintech companies that pride themselves on a great developer experience, this will play to their advantage. Fintechs are already prioritizing the creation of developer sandboxes to let customers “try before you buy,” and are even open sourcing parts of their solutions. The developer-buyer would, of course, prefer to get a sense of how the product works prior to full implementation. But isn’t this strategy also better for everyone?

For larger financial institutions selling their infrastructure, appealing to the developer will be a new muscle that may require improvements in product architecture (including up-to-date documentation!). Still, it’s clear these institutions are recognizing the need to influence the developer, as well. 

2024 年,开发人员将成为购买金融服务基础设施的最重要影响因素之一。


在各种规模的金融服务公司中,开发者作为买家的崛起有利于新进入者;对于那些以出色的开发者体验为傲的金融科技公司来说,这将对他们有利。金融科技公司已经将创建开发者沙盒作为优先事项,让客户 "先试后买",甚至将其解决方案的部分内容开源。当然,开发者-买家更希望在全面实施之前了解产品的运行情况。但这种策略不是对所有人都更好吗?


25)Tech Helps Community and Regional Banks Compete


In the wake of SVB and First Republic, community and regional banks are facing significant regulatory pressure and margin compression from a higher rate environment. Despite challenges, these institutions remain vital for the health and vibrancy of our local economies and small businesses. My hope is that we’ll see fintech companies step up and provide needed tools and technology to help this ecosystem of banks compete with larger institutions, manage balance sheet risks effectively, and better serve their clients. The time is now! 

在 SVB 和 First Republic 倒闭之后,社区银行和地区银行正面临着巨大的监管压力和高利率环境对利润的压缩。尽管面临挑战,但这些机构对于我们当地经济和小企业的健康与活力仍然至关重要。我希望我们能看到金融科技公司挺身而出,提供所需的工具和技术,帮助这些银行生态系统与大型机构竞争,有效管理资产负债表风险,并更好地为客户服务。时不我待!

26)Financial Professional Services Get Supercharged by Software


The professional services work of financial services — the accountant, the tax adviser, the wealth manager, the investment banker — will change. These professions typically involve researching and applying learned expertise, as well as client management. Historically, software has mainly assisted in tracking workflows, with only some analytical tools (e.g. transaction categorization in accounting). With the advancement of generative AI and LLMs, more of the work can be automated, including administrative tasks, the research process (collecting and ingesting data, searching for information), summarizing and surfacing insights, and report generation. A tax advisor can more easily look through precedent to answer a question, an accountant can automatically generate a financial statement, a wealth advisor can scenario plan across a broader set of data.

Software may eventually automate the task entirely, but for now the skill set for humans in the loop will shift toward specialized expertise, reviewing the generated work, and client-facing engagement. As always, it’ll be a race between distribution and innovation: Incumbents that have already established relationships with financial professionals will need to incorporate AI into their software, while startups with modern software capabilities will need to find and build trust among new customers.

金融服务的专业服务工作--会计师、税务顾问、财富经理、投资银行家--将发生变化。这些职业通常涉及研究和应用所学专业知识以及客户管理。从历史上看,软件主要协助跟踪工作流程,只有一些分析工具(如会计中的交易分类)。随着生成式人工智能和 LLM 的发展,更多的工作可以实现自动化,包括管理任务、研究过程(收集和摄取数据、搜索信息)、总结和浮现见解以及生成报告。税务顾问可以更轻松地通过查阅先例来回答问题,会计师可以自动生成财务报表,财富顾问可以通过更广泛的数据集进行情景规划。

软件最终可能会实现任务的完全自动化,但就目前而言,环路中的人类技能组合将转向专业知识、审查生成的工作以及面向客户的参与。与以往一样,这将是一场分销与创新之间的竞赛: 已经与金融专业人士建立关系的现有公司需要将人工智能融入其软件,而拥有现代软件能力的初创公司则需要在新客户中寻找并建立信任。

27)LLMs Capture New “Foundational Customer Units”

大语言模型捕捉新的 "基础客户群"

Operating systems are incredibly valuable — they own what we call the foundational customer unit (FCU). Historically, certain types of unstructured data have been challenging for operating systems to ingest. For example, Vertafore or Applied Systems in insurance have struggled to expand outside of tracking policies once they are written, since collecting data earlier in the process requires ingesting information that lives in email, PDFs, or spreadsheets. In 2024, startups leveraging LLMs will capture data that has been challenging for incumbent operating systems to collect, while automatically tagging and storing it. If these startups capture FCUs upstream from traditional platforms, we may see segments that have been served by software oligopolies shift into a new era. 

操作系统具有惊人的价值--它们拥有我们所说的基础客户单元(FCU)。从历史上看,对于操作系统来说,某些类型的非结构化数据的摄取具有挑战性。例如,保险业的 Vertafore 或 Applied Systems 一直在努力拓展保单撰写后的跟踪之外的业务,因为在流程的早期收集数据需要摄取电子邮件、PDF 或电子表格中的信息。2024 年,利用 LLM 的初创公司将捕获现有操作系统难以收集的数据,同时自动标记和存储这些数据。如果这些初创公司能从传统平台的上游捕捉到FCU,那么我们可能会看到一直由软件寡头垄断的细分市场进入一个新时代。

28)New Tools for Banking and Trading


In 2024, we’ll see ambitious founders tackle some of the messiest problems that financial institutions have to offer. 

Though the investment banking and trading services market produces nearly $350 billion of annual revenues globally, it still largely relies on systems and software built on-prem in the 1980s. While banks have started to buy cloud-based solutions to serve various horizontal needs — Salesforce for CRM, Azure for cloud computing, Databricks for lakehouse architecture — the tools vertically deployed in the banking and trading businesses to model risk, confirm/settle/clear trades, and record client orders are often manual (Excel), outdated, or both! At the same time, the buying behavior at these institutions is changing. Willingness to try new tools is at an all time high. 

2024 年,我们将看到雄心勃勃的创始人解决金融机构面临的一些最棘手的问题。

尽管投资银行和交易服务市场在全球范围内创造了近3500亿美元的年收入,但它在很大程度上仍然依赖于上世纪80年代在内部构建的系统和软件。虽然银行已经开始购买基于云的解决方案来满足各种横向需求--Salesforce 用于 CRM、Azure 用于云计算、Databricks 用于湖居架构--但银行和交易业务中纵向部署的用于风险建模、确认/结算/清算交易以及记录客户订单的工具往往是手动(Excel)、过时或两者兼而有之!与此同时,这些机构的购买行为也在发生变化。尝试新工具的意愿达到了历史最高点。

29)AI Will Push Latin American SMBs to Go Digital


Many merchants in Latin America use WhatsApp for customer service and support. These interactions predominantly involve tasks in which consumers expect quick responses, such as quoting, scheduling, and logistics. Currently, merchant response times can vary greatly depending on representatives’ availability and the complexity of the request. AI assistants could significantly streamline these time-consuming tasks, unlocking value for merchants and consumers. 

One recent example: this year, Nissan developed chatbots on WhatsApp that direct customers in Brazil to one of its nearby car dealerships. Nissan estimates that 30 to 40% of its sales leads in Brazil are now generated via WhatsApp, and the company’s average response time dropped from 30 minutes to a few seconds. 

However, though such early inroads are promising, structural challenges remain before the technology can be widely adopted by Latin American SMBs. In Brazil, for instance, over 40% of SMBs still rely on pen-and-paper operations. As automated customer service becomes increasingly prevalent and consumers grow accustomed to these prompt interactions, businesses will be compelled to digitize their workflows. There’s an opportunity for startups to facilitate this digital shift in the region. 

拉丁美洲的许多商家使用 WhatsApp 提供客户服务和支持。这些互动主要涉及消费者希望得到快速响应的任务,如报价、日程安排和物流。目前,商家的响应时间会因客服代表的空闲时间和请求的复杂程度而有很大差异。人工智能助手可以大大简化这些耗时的任务,为商家和消费者创造价值。

最近的一个例子是:今年,日产汽车公司在 WhatsApp 上开发了聊天机器人,可以将巴西的客户引导到附近的一家汽车经销商。据日产公司估计,它在巴西的销售线索中有 30% 至 40% 现在都是通过 WhatsApp 生成的,公司的平均响应时间也从 30 分钟缩短到了几秒钟。

然而,尽管这些早期进展令人充满希望,但在拉美中小企业广泛采用该技术之前,仍然存在结构性挑战。例如,在巴西,超过 40% 的中小企业仍然依赖纸笔操作。随着自动化客户服务的日益普及,消费者逐渐习惯于这些及时的互动,企业将不得不将其工作流程数字化。初创企业有机会在该地区推动这一数字化转变。

30)AI Will Be the Key to Higher ROE

人工智能将成为提高 ROE 的关键

In the coming year, we’ll begin to see financial institutions adopt AI-native applications across a variety of operational workflows (and we’ll likely be surprised at how big of an economic impact the technology has on their P&L). While the opportunity set spans both revenue generating and middle- to back-office functions, adoption in 2024 will be focused on use cases throughout engineering, procurement, legal, compliance, and risk management.

未来一年,我们将看到金融机构开始在各种业务工作流程中采用人工智能原生应用(我们可能会惊讶于该技术对其损益产生的经济影响有多大)。虽然机会集横跨创收和中后台功能,但 2024 年的应用将集中在工程、采购、法律、合规和风险管理等方面的用例。

Games | 游戏

31)Gaming: Where Alpha Geek Technologies Get Their Initial Product Market Fit

游戏: 阿尔法极客科技公司的初始产品市场契合点在哪里

There are many new technologies on the cusp — AI, VR/AR, web3 — and embracing video games will be a key aspect of their success. These new technologies will change gaming, but will also be changed by gaming. For generative AI, the next stage after text and images will be 3D and video. Combined with audio, interactivity, and other facets, eventually games will cost 1/1,000th what they traditionally did to develop and will enable consumers to create their own gaming experiences.

VR/AR has found the most product/market fit by targeting kids and teens with multiplayer gaming experiences. The next generation of headsets would do well to double down, acquiring millions of consumers in the process, rather than trying to jump to productivity tools where demand is low. For web3, each wave has been propelled by a major use case — NFTs, DeFi, etc. The next major consumer wave will come from fun, mainstream games that introduce web3 as a way for gamers to engage with the virtual items they buy.

人工智能、VR/AR、web3 等许多新技术正处于风口浪尖,而拥抱电子游戏将是这些技术取得成功的关键因素。这些新技术将改变游戏,但也将被游戏所改变。对于生成式人工智能而言,继文字和图像之后的下一个阶段将是 3D 和视频。结合音频、互动性和其他方面,最终游戏的开发成本将只有传统游戏的千分之一,消费者将能够创造自己的游戏体验。

VR/AR 通过针对儿童和青少年的多人游戏体验,找到了最适合的产品/市场。下一代头显最好能加倍努力,在这一过程中获得数百万消费者,而不是试图跳到需求较低的生产力工具上。对于网络3来说,每一次浪潮都是由一个主要使用案例推动的--NFT、DeFi等。下一个主要的消费浪潮将来自有趣的主流游戏,这些游戏将 web3 作为游戏玩家与他们购买的虚拟物品互动的一种方式。

32)AI-First Games That Never End


In 2024, we’ll see the first cohort of AI-first games from creators leveraging large models to enable novel game systems and mechanics. While much of the early discourse on generative AI in games has focused on how AI can make game creators more efficient, I believe the largest opportunity long-term is in leveraging AI to reinvent the nature of the games entirely — creating never-ending games that engage and retain users for a long time.

Generative agents powered by large language models will create incredibly lifelike companions and emergent social behavior, turbo-charging games with non-player characters (NPCs). Personalized character-builders and narrative systems will enable every player to have unique, personalized playthroughs of their favorite games. Game worlds themselves will no longer be rendered, but generated at run-time using neural networks. And new player onboarding will be reinvented. Every game will be designed around an AI copilot with the mantra “good alone, great with AI, and best with friends.”

2024 年,我们将看到第一批人工智能游戏,它们来自利用大型模型实现新颖游戏系统和机制的创作者。虽然早期关于游戏中的生成式人工智能的讨论大多集中在人工智能如何提高游戏创作者的效率上,但我认为最大的长期机遇在于利用人工智能完全重塑游戏的本质--创造出能够吸引和留住用户很长时间的永无止境的游戏。

由大型语言模型驱动的生成代理将创造出栩栩如生的同伴和新出现的社交行为,为拥有非玩家角色(NPC)的游戏注入新的活力。个性化的角色创建器和叙事系统将使每个玩家都能在自己喜爱的游戏中体验独特的个性化游戏。游戏世界本身将不再是渲染出来的,而是在运行时利用神经网络生成的。新玩家的加入也将被重塑。每款游戏都将围绕人工智能副驾驶员进行设计,其口号是 "单独玩很好,与人工智能一起玩很棒,与朋友一起玩最好"。

33)Games Become “Everything Simulators”

游戏成为 "万物模拟器"

Games are simulations that satiate our base biological primitives: for collection (Pokemon), for predator/prey (tag), for nurturing (AdoptMe), for exploration (Minecraft). Game engines facilitate the laws of the simulation. What game engines could not efficiently simulate, however, were the complex and emergent nature of human thoughts, actions, words, and goals. That is, until recently.

Breakthroughs in LLMs and agentic frameworks now make it possible to have realistic characters in games with believable goals, actions, and dialogue. This gives the game designer a new tool in their toolbelt: the ability to simulate social dynamics. In 2024, expect to see games where the moment-to-moment involves coercion, deception, flirting, allying, leadership, peer pressure, influence, morality… you name it. Simulation tools never before possible will now be generally available.

Every biological primitive will be fair game — your innate human desire to socialize, cooperate, to find love lies at the whims of the simulation.


现在,LLM 和代理框架的突破使得游戏中的角色具有真实可信的目标、行动和对话成为可能。这为游戏设计师提供了新的工具:模拟社会动态的能力。2024 年,我们有望在游戏中看到胁迫、欺骗、调情、结盟、领导力、同侪压力、影响力、道德......你说得出来。以前从未有过的模拟工具现在将普遍可用。


34)From Chatbots to Avatars: AI Companions Go 3D

从聊天机器人到头像: 人工智能伴侣走向 3D

If 2023 was the year of AI companions we text with, next year those relationships will come to life through 3D avatars we talk to in real-time, verbally. AI companion apps like Character AI are already seeing millions of monthly active users engage with chatbots like virtual Elon Musk, Super Mario, or a psychologist. In the coming year, these conversations will feel as natural as a FaceTime conversation. With lower latency responses, text-to-speech advances, and audio-driven facial animation, our conversations with AI companions will feel increasingly perceptive, present, and personalized. Entertainment will continue to shift from a passive to active experience, and the lines will blur between linear TV and interactive games. 

如果说 2023 年是我们与人工智能伴侣发短信的一年,那么明年,这些关系将通过我们与之实时交谈的 3D 头像变得栩栩如生。目前,Character AI 等人工智能伴侣应用程序已经拥有数百万月活跃用户,他们与虚拟埃隆-马斯克(Elon Musk)、超级马里奥(Super Mario)或心理学家等聊天机器人进行了互动。未来一年,这些对话将像 FaceTime 对话一样自然。随着响应延迟的降低、文本到语音技术的进步以及音频驱动的面部动画,我们与人工智能伙伴的对话将越来越具有感知力、存在感和个性化。娱乐将继续从被动体验向主动体验转变,线性电视和互动游戏之间的界限将变得越来越模糊。

35)The Next Disney is a Game Company


There’s been a lot of commentary recently on Disney getting back into video games, but I believe the next Disney will be a video game company. 

$188BExpected games revenue in 2023. By comparison, the global movie box office is expected to make $34.5B.

2023 was a banner year for games in film and TV. The Super Mario Bros. Movie was topped only by Barbie, and The Last of Us series was the second best performing HBO series of the last decade. Though Hollywood has typically punched above its weight culturally, the global games market has never been stronger. Games are expected to gross $188 billion worldwide this year, while the global box office is only projected to reach $34.5 billion.  

As younger games-native generations play Roblox, Fortnite, Clash of Clans and Valorant, they are increasingly turning to games as their IP of choice. The reason? Games provide the deepest stories and worlds, they’re interactive instead of passive, and they offer engaging social experiences. Studios are embracing AI, which is turbocharging game creation.

When Riot Games launched the series Arcane, based on League of Legends, it became Netflix’s #1 watched and top rated series. Companies like Riot, Epic, Supercell, and new next generation game companies are poised to become the next entertainment juggernauts with games replacing films as the core of the “next Disney.” This is already happening without the mainstream noticing and will accelerate in 2024. 


2023 年的游戏收入预计为 188 亿美元。相比之下,全球电影票房预计收入为 345 亿美元。

2023 年是游戏在影视领域大放异彩的一年。超级马里奥兄弟电影》的票房仅次于《芭比娃娃》,而《最后的我们》系列则是 HBO 近十年来票房第二好的剧集。虽然好莱坞在文化上通常都是一枝独秀,但全球游戏市场却从未如此强劲。预计今年全球游戏总收入将达到 1880 亿美元,而全球票房预计仅为 345 亿美元。

随着年轻一代游戏原住民开始玩《Roblox》、《堡垒之夜》、《部落冲突》和《Valorant》,他们越来越多地将游戏作为自己的首选 IP。原因何在?游戏提供了最深刻的故事和世界,它们是互动的,而不是被动的,还提供了引人入胜的社交体验。工作室正在拥抱人工智能,而人工智能正在加速游戏创作。

当 Riot Games 推出以《英雄联盟》为蓝本的系列剧《Arcane》时,它成为 Netflix 的收视冠军和收视率冠军。像 Riot、Epic、Supercell 这样的公司,以及新的下一代游戏公司,都将成为下一个娱乐巨头,游戏将取代电影,成为 "下一个迪斯尼 "的核心。这种情况已经在不知不觉中发生,并将在 2024 年加速发展。

36)Anime Games Go Mainstream


Anime has become one of the highest grossing genres, when calculated by average revenue per user. In 2022, Mihoyo grossed over $3.8 billion from releases like Genshin Impact and Honkai: Star Rail. Earlier this year, Nintendo launched Zelda: Tears of the Kingdom. In 2024, I expect anime’s momentum to continue. 

Anime is a uniquely accessible art style for kids and adults, and the medium allows for many different story archetypes. There are adventure components, romance hooks, and social loops within anime games that lead to deep player engagement. Genshin set the new standard for a fully cross-platform, performant game across multiple device types, GPUs, and frameworks. Many passionate developers are now building new experiences for players everywhere. 

如果按每个用户的平均收入计算,动漫已成为收入最高的类型之一。2022 年,《源氏冲击》和《Honkai: Star Rail》等作品使美宝欧的总收入超过 38 亿美元。今年早些时候,任天堂推出了《塞尔达:王国之泪》。2024 年,我预计动漫的发展势头将继续保持。

动漫是一种独特的艺术风格,儿童和成人都能接受,而且这种媒体允许有许多不同的故事原型。动漫游戏中的冒险元素、浪漫情节和社交循环都能让玩家深度参与其中。元心》为完全跨平台、跨多种设备类型、GPU 和框架的高性能游戏设定了新标准。现在,许多充满激情的开发人员正在为世界各地的玩家打造全新的游戏体验。

37)A New Generation of UGC Game Developers Emerges

新一代 UGC 游戏开发者崛起

Increasing production budgets and heightened player expectations have made it difficult to succeed as a game developer. But new user-generated content (UGC) platforms and AI-powered creator tools show promise in breaking down those barriers. 

In Q1 2023, Roblox developers earned $182 million, an estimated 17% increase from 2022. Epic has also started financially supporting Fortnite Creative developers, with projected payouts surpassing $100 million in 2023. As competition among UGC platforms grows, developers stand to benefit from greater incentives. Notably, Meta’s Horizon Worlds expanded to mobile in 2023.

Alongside better financial backing, UGC game developers now also have access to more powerful tooling driven by generative AI. (Epic has been publicly supportive of such technology and Roblox has already announced some gen AI tools.) Combined, these two factors are likely to unleash millions of new creators next year. 


2023 年第一季度,Roblox 开发者的收入为 1.82 亿美元,预计比 2022 年增长 17%。Epic 也开始为《堡垒之夜》创意开发者提供资金支持,预计 2023 年的报酬将超过 1 亿美元。随着 UGC 平台之间的竞争日益激烈,开发者将从更大的激励机制中获益。值得注意的是,Meta 的《地平线世界》在 2023 年扩展到了移动平台。

除了更好的资金支持外,UGC 游戏开发者现在还可以使用由生成式人工智能驱动的更强大的工具。(Epic已经公开支持此类技术,而Roblox也已经公布了一些生成式人工智能工具)。这两个因素结合在一起,很可能在明年释放出数百万新的创作者。

38)The Minecraft Generation Ignites the Survival-Crafting Genre


Minecraft will celebrate its 13th anniversary next year, marking a milestone as a generation of players nurtured on “crafting” games transitions into the adult gaming category. But there’s no clear game or experience that seems to support their interests. On one end of the spectrum is Rust, a fiercely competitive “crafting-survival” game-as-a-service (GaaS) behemoth first published in 2013. A more friendly offering is Valheim, launched in 2021, which smashed commercial forecasts and spurred a wave of developers to emulate its success. Many are now aspiring to create a GaaS rendition of this genre blend. The Valheim zeal just might catalyze the creation of a new, billion-dollar IP that resonates with the maturing preferences of the “Minecraft Generation.” 

明年,《我的世界》(Minecraft)将迎来 13 周年纪念,这标志着在 "手工制作 "游戏中培养起来的一代玩家正在向成人游戏过渡,这是一个里程碑。但目前似乎还没有一款明确的游戏或体验能支持他们的兴趣。Rust》是一款竞争激烈的 "制作-生存 "类游戏,于2013年首次发布。2021年推出的《瓦尔海姆》(Valheim)则更为友好,它打破了商业预测,并引发了一波开发者效仿其成功的浪潮。现在,许多人都希望创造出这种类型混合的 GaaS 版本。Valheim》的热潮可能会催生一个价值数十亿美元的新 IP,与 "Minecraft 一代 "日渐成熟的喜好产生共鸣。

Growth-Stage Tech | 成长期技术

39)New Modes of Storytelling, Beyond Text-Based Chat


Character.AI CEO Noam Shazeer has called entertainment “AGI’s first use case” — the ability to use AI to tell stories across text, audio, and visual formats has continued to improve at a rapid rate. In the coming year, AI will advance beyond text-based chat to multimodal models. Layering in personalization and fine-tuning by users will deepen the way we interact with AI and enable even more exciting, entertaining, and engaging experiences. It’s up to startups to create these new storytelling formats.  

Character.AI 首席执行官诺姆-沙泽尔(Noam Shazeer)将娱乐业称为 "AGI 的第一个用例"--利用人工智能讲述文本、音频和视觉形式故事的能力在不断快速提高。未来一年,人工智能将从基于文本的聊天发展到多模态模式。个性化的分层和用户的微调将深化我们与人工智能的互动方式,并带来更多令人兴奋、有趣和引人入胜的体验。现在就看初创企业如何创造这些新的故事形式了。

40)CRM Will Be Powered by AI Data Collection


Sales rep data is the atomic unit of the go-to-market organization, and bad data is a problem for almost every GTM leader. No matter how many tools or plugins you throw at your CRM platform, you still have the same fundamental problem: your reps need to input accurate data. If reps input crummy data, you get crummy results.

Though we’ve seen some sales tech companies experiment with generative AI around the edges, the next generation of sales tech will use generative AI to tackle this core data problem head-on. Instead of relying on reps’ recollections or interpretations of customer meetings, these AI-native companies will be built on ground-source data, automatically captured or generated by AI, from actual customer interactions like meeting notes, emails, and call recordings.

We’ll likely see massive bottom-up adoption of these sales tools because they’ll drive significant productivity improvements. Eventually, these new AI-native sales tech companies could pave the way for a fully AI-native CRM.

销售代表数据是市场推广组织的原子单元,而糟糕的数据几乎是每个 GTM 领导者都会遇到的问题。无论你在 CRM 平台上使用多少工具或插件,你仍然会遇到同样的基本问题:你的销售代表需要输入准确的数据。如果销售代表输入的数据不准确,你就会得到不准确的结果。


我们很可能会看到这些销售工具自下而上地被大量采用,因为它们将推动生产力的显著提高。最终,这些新的人工智能原生销售技术公司可能会为完全人工智能原生的 CRM 铺平道路。

41)The Consumer AI Battleground Moves from Model to UX


Overheard in 2023: “A minute not spent building the model is a minute wasted. Build the best model, and users will come.”

The most popular consumer AI companies thus far have been producers of their own models, such as ChatGPT, Character, Bard, and Midjourney. Differentiation has come from being the best model in their domain: Midjourney with images, Character on entertainment, ChatGPT on overall text. UX has largely been determined by the fastest way to get models into the hands of users.

But thanks to a combination of factors — the likelihood that chip shortages ease, the availability of most foundation models via API, and increasingly powerful open source models — the groundwork is there for breakout consumer apps to be built on someone else’s model.

In 2024, consumer AI apps will break out by delivering the best user experience around unique use cases, less on model performance alone. I’m particularly excited for consumer AI apps that figure out how to include shared experiences and a multiplayer mode, aggregate multiple models into a single interface, or build more focused solutions where workflow and process drive value. LLMs can be a source of differentiation. Today, they might provide a first mover advantage, but old-fashioned moats like network effects, high switching costs, scale, and brand are still what will likely win long-term.

在 2023 年听到的 "没有建立模型的一分钟就是浪费的一分钟。建立最好的模型,用户就会来。"

迄今为止,最受欢迎的消费人工智能公司都是自己模型的生产者,如 ChatGPT、Character、Bard 和 Midjourney。差异化来自于成为各自领域的最佳模型: Midjourney是图像模型,Character是娱乐模型,ChatGPT则是整体文本模型。用户体验在很大程度上取决于以最快的方式将模型送到用户手中。


2024 年,消费类人工智能应用将通过围绕独特的使用案例提供最佳用户体验而实现突围,而不是仅仅依靠模型性能。我尤其期待消费类人工智能应用能够想出办法,将共享体验和多人游戏模式纳入其中,将多个模型聚合到一个界面中,或者构建更有针对性的解决方案,让工作流程和过程推动价值的实现。LLM 可以成为差异化的源泉。如今,它们可能会提供先发优势,但网络效应、高转换成本、规模和品牌等老式护城河仍可能是长期制胜法宝。

Infra + Enterprise | 基础设施 + 企业

42)AI Interpretability


Interpretability, which is just a complex way of saying “reverse engineering” AI models, will be big in 2024. Over the last few years, AI has been dominated by scaling — a quest to see what was possible if you threw a ton of compute and data at training these models. Now, as these models begin to be deployed in real world situations, the big question is: why? Why do these models say the things they do? Why do some prompts produce better results than others? And, perhaps most importantly, how do we control them?

可解释性是 "逆向工程 "人工智能模型的一种复杂说法,将在 2024 年大行其道。在过去几年里,人工智能一直被规模化所主导--人们想知道,如果投入大量计算和数据来训练这些模型,会有什么可能。现在,随着这些模型开始部署到现实世界中,最大的问题是:为什么?为什么这些模型会说出这样的话?为什么有些提示会比其他提示产生更好的结果?也许最重要的是,我们该如何控制它们?

43)Creativity Reimagined


Creativity is the most intrinsic mode of human expression, but ideas are abstract — expressing them clearly takes time and skill. Generative AI has created a path to truly democratize the means to create. From writing to painting to filmmaking, what used to take a team months to perfect now takes minutes, if not seconds. It gives everyone, skilled or not, the ability to create.

The creative workflow has been fundamentally revamped. Prototyping and ideation are incredibly interactive. Writer’s block can now be solved by iterating with a copilot, and artistic skill sets can be honed through iteration, rather than repetition. In this new paradigm, a new set of tools are emerging that allow us to express our creativity in a multidimensional way. The key element here will be learning to compose in different modalities, including text, visual form, and audio. In 2024, these new AI playgrounds will make creative expression even more widely accessible.


创意工作流程已经发生了根本性的改变。原型设计和构思具有令人难以置信的互动性。写作障碍现在可以通过与副驾驶员反复讨论来解决,艺术技能可以通过反复而不是重复来磨练。在这种新模式下,一套新的工具正在出现,使我们能够以多维的方式表达自己的创造力。这里的关键因素将是学会以不同的方式进行创作,包括文字、视觉形式和音频。2024 年,这些新的人工智能游乐场将使创意表达更加普及。

44)B2B AI Products Embedded in Your Workflow

嵌入工作流程的 B2B 人工智能产品

Generative AI and other technologies have the potential to automate70%of the time employees spend working today.

In 2024, I’m optimistic we’ll see AI-native products become more embedded into workflows, performing tasks like proactively leaving comments, updating records, and completing action items after a simple approval from the user. We’re already seeing workflow-native AI products take more direct action on users’ behalf. Instead of waiting for the user to query a long document for relevant information, for example, an AI tool can proactively flag key sections.

In tandem, within B2B products I expect to see a shift away from the chat UX. Chat has been conducive in demonstrating the usefulness of LLMs, but the prompt interface ultimately disconnects the user’s workflow. In 2024, I believe we’ll see innovative AI products that are designed to meet users where they already are.  

生成式人工智能和其他技术有可能将目前员工工作时间的 70% 自动化。

在 2024 年,我很乐观地看到,人工智能原生产品将更多地嵌入到工作流程中,执行主动留言、更新记录以及在用户简单批准后完成操作项目等任务。我们已经看到工作流原生人工智能产品代表用户采取更直接的行动。例如,人工智能工具可以主动标记关键部分,而不是等待用户查询长文档中的相关信息。

与此同时,在 B2B 产品中,我预计聊天用户体验也将发生转变。聊天有助于展示 LLM 的实用性,但提示界面最终会切断用户的工作流程。2024 年,我相信我们将看到创新的人工智能产品,这些产品旨在满足用户的需求。

45)The New “Division of Labor”

新的 "分工

As technology evolves, it will enhance our creative and problem-solving abilities.

The big question: What will the new means of creation look like?


最大的问题是 新的创造手段会是什么样子?

46)LLMs Advance Robotic Process Automation Systems


In 2024, I’ll be excited to see the market for LLM-powered robotic process automation (RPA) companies take off. Today, enterprises often have manual processes done on legacy software systems that are too difficult to rip and replace or to build deep integrations into. In these situations, RPA — where small “bots” are deployed to automate repetitive tasks, such as data entry — is currently the best solution. However, RPA is often still very manual and frequently breaks; it often requires a lot of custom implementation and services to stand up.

With LLMs, there is an opportunity to build a more intelligent RPA system that can contextually understand the inputs and actions it’s taking and will be able to dynamically adjust to create a more robust solution. It’s likely there will be multiple verticalized solutions tailored to specific types of automation tasks — whether for the finance org, processing invoices, or for support org, responding to customer service inquiries — as buyers purchase the solution most tailored to their workflows and needs. 

2024 年,我很高兴看到由 LLM 驱动的机器人流程自动化(RPA)公司市场起飞。如今,企业的人工流程往往是在传统软件系统上完成的,而这些系统很难被改造和替换,也很难进行深度集成。在这种情况下,RPA(部署小型 "机器人 "来自动完成重复性任务,如数据录入)是目前最好的解决方案。但是,RPA 通常仍然是非常人工的,而且经常出现故障;它通常需要大量的定制实施和服务才能建立起来。

有了 LLM,我们就有机会建立一个更智能的 RPA 系统,该系统可以根据上下文理解输入和执行的操作,并能够动态调整,从而创建一个更强大的解决方案。很可能会出现针对特定类型自动化任务定制的多种垂直化解决方案--无论是为财务部门处理发票,还是为支持部门响应客户服务咨询--因为买家会购买最适合其工作流程和需求的解决方案。

本文》有 0 条评论