AI资讯 2025/7/18
AI 日报
AI内容摘要
AI行业喜忧参半。OpenAI智能体可自主操作网页，AI视频与编程助手也涌现，极大提升效率。
Anthropic估值飙升，但面临大量版权诉讼，凸显技术进步与版权保护的冲突。
AI正渗透生活各领域，如数字伴侣，带来便利的同时也引发对人机关系的深刻思考。
Today’s AI News
The AI industry has truly been a “mixed bag” recently: on one side, there are exciting technological breakthroughs, with AI becoming increasingly like a capable “efficient manager”; on the other, there are intense “legal battles”, as AI companies face lawsuits for “consuming” data they shouldn’t have.
First, the good news! OpenAI recently unleashed a powerful weapon – ChatGPT Agent. This isn’t your usual conversational AI. It’s like giving ChatGPT a “body”, equipping it with its own “virtual computer”. From now on, your AI assistant won’t just “talk”; it can autonomously browse the internet, view web pages, click buttons, fill out forms, and even help you shop, analyze data, create PPTs, or plan travel itineraries. You just give it a single command, and it acts like a super “toolset”, figuring out how to use various “tools” to get things done. For instance, you could ask it to find and summarize the latest news about meeting clients, or to design a Japanese breakfast menu and directly order the ingredients, or even analyze three competing companies and automatically generate a presentation. Doesn’t that sound like having a tireless “worker”? Moreover, OpenAI emphasizes that the “user is always in control”, and any “high-risk” operations involving privacy or payments require your “explicit authorization”. You can stop it anytime, ensuring complete security. This feature has performed exceptionally well in various tests and may even become the foundation for more powerful AIs like GPT-5, potentially even integrating payment systems. With this, couldn’t you happily go enjoy a coffee while leaving tedious data research and report generation to AI?
However, AI’s rapid advance has also brought considerable trouble. On the other side, AI company Anthropic has found itself in a big mess, being sued by a group of “writers”. This isn’t a small skirmish; it’s a “class-action lawsuit”! The writers accuse Anthropic of “stolen millions of copyrighted books” to train its AI, much like the “pirate music haven” Napster of yesteryear, directly “transporting” data from “pirated libraries”. Although Anthropic argued that it had legally purchased some books for training (which might be considered “fair use”), the judge stated that Anthropic still faces a “separate trial” for the “allegedly pirated books”. This isn’t an isolated case; Reddit and Universal Music have also successively sued Anthropic for “copyright infringement”.
This reflects the fierce collision between “AI technology development” and “copyright protection”. When AI “devours” massive amounts of data like a bottomless pit to learn and grow, who pays the original creators of that data? Should AI companies be allowed to grow wildly like headless chickens, or should regulations be established for the “creative industries” to ensure creators’ rights? This is not just a legal issue but also a profound question about our social ethics and the future of creative models.
So, you see, the future of AI is full of infinite possibilities, capable of freeing us from tedious work and creating more value; but at the same time, it also demands that we sit down and seriously consider how to uphold the bottom line of fairness and justice amidst the rapid rush of technological innovation. Where will this “tug-of-war” between “artificial intelligence” and “copyright” lead in the future? This is worth our continued attention and contemplation.
Is AI evolving so fast that it feels like we’ve stepped into a time machine? Recently, three major “game-changers” have been exposed, making us feel directly that AI is upgrading from a “chat king” to a “multi-faceted” player, even starting to compete with “Hollywood directors” and “super secretaries” for jobs!
First up is Moonshot AI’s Kimi, whose Kimi Playground has completely revolutionized our perception of AI: this thing no longer just chats with you; it can actually “take action”! Imagine, you ask it “What’s the weather tomorrow?”, and it no longer replies with “I don’t know” but instead “calls upon” a “weather query tool” and instantly gives you accurate information. This is like equipping AI with a “Swiss Army knife” “toolset”, packed with various “practical skills”: booking flights and hotels, planning travel routes, analyzing complex data… Previously, these tasks required you to painstakingly run several apps yourself, but now Kimi K2 can handle them “one-stop solution”, acting as your “super personal assistant”! For “developers”, this is even more of a “playground”, allowing them to easily add “skill packs” to AI, with even “third-party tools” like ModelScope MCP able to be “seamlessly integrated”. In the future, AI will be like a “super apprentice” capable of learning new skills infinitely, which is quite exciting to imagine!
Meanwhile, Lightricks dropped a “bombshell” – the “open-source AI video generation model LTX-Video 13B”! This thing essentially brings the “Hollywood film production workflow” to your computer. How powerful is it? First, it’s “30 times faster” than comparable models! Previously, AI video creation was incredibly slow, but now it’s like a “Ferrari”, and even more incredibly, it can run on your home “consumer-grade graphics cards”, no expensive servers needed! The core technology is “multi-scale rendering”, much like a painter first sketching and then refining details, so the generated videos are both “smooth” and “clear”. Most importantly, it can generate “native videos” “up to 60 seconds” in length in one go! Think about it, previous AIs could only give you a few-second clip, but now it directly gives you the beginning of a “micro-movie”. What’s even more amazing is that you can control it like a “true director”: adjusting character movements via “key frames”, controlling camera zooms and pans, and even performing “video style transfer”. This means you can not only “one-click blockbuster generation” but also “personally direct blockbusters”, and the era where everyone can be Ang Lee might truly be here!
And our everyday ChatGPT is quietly getting “lazier”—not that it’s getting lazy, but it’s making us “lazier”! The macOS desktop application has launched a “recording mode”. Now, for meetings, brainstorms, or just to jot down fleeting inspirations, you can simply “talk” to ChatGPT! It will “real-time transcription” what you say into text, then “automatically summarize key points”, and also “save” these “crystallizations of wisdom.” What’s even more magical is that you can ask it anytime: “What did we discuss in the last meeting?” It can “instantly recall” and help you turn the recording summary into “meeting minutes”, “email drafts”, “project plans”, or even “directly write code”! This function is truly like having a combined “super shorthand typist” and “all-around secretary”, not only helping you “organize thoughts” but also facilitating “efficient output”. As for “privacy”? It’s been considered; “audio recordings are immediately deleted”, allowing you to use it with greater peace of mind.
These three breakthroughs show us that AI is no longer just a single “chatbot” or “image generator”; it is becoming a “multi-faceted” entity, capable of “speak”, “act”, and “create with thought”. AI is moving from “nice-to-have” to “essential support”, gradually integrating into all aspects of our work and life, becoming our indispensable “efficiency multiplier” and “creativity catalyst”. This undoubtedly greatly “liberates human hands and minds”, allowing us to focus our energy on more creative and strategic work. But at the same time, we must also consider: as AI becomes increasingly “capable” and even able to handle so many things, how should we define our human value? Where are the boundaries of “human-AI collaboration”? This is perhaps the most profound “philosophical contemplation” left to us by the AI era.
Hey, tech enthusiasts! The AI world is buzzing today, from “100 billion valuation” giants to the birth of “3D AI girlfriends” – it’s truly a “wild ride”!
First off, AI star player Anthropic’s “valuation” has soared like a rocket, directly breaking the "$100 billion" mark, doubling in just a few months! This surge is no joke; their “annualized revenue” has also exploded, hitting “over $4 billion”, quadrupling. A major contributor to this success is their “AI programming” assistant, Claude Code, whose “download volume” has “doubled weekly”, contributing “over $100 million in annual revenue” to the company. However, making money isn’t always simple; while Anthropic directly selling models yields a “high gross margin”, selling through Amazon and Google cloud platforms is actually “losing money”! It’s a striking reminder that while developing AI burns a lot of “cash” (they anticipate “burning $3 billion” this year), as long as they can generate “revenue”, “capital” will still pursue them relentlessly!
Speaking of “AI programming”, another explosive piece of news has emerged—the well-known “AI programming tool” Windsurf platform has finally brought back “Anthropic’s Claude Sonnet 4 model”! This model was previously “shelved” by Anthropic, rumored because OpenAI wanted to “acquire Windsurf”, leading “competitors” Anthropic to be “unhappy” and withhold “direct authorization”. Now, Sonnet 4 has “returned to Windsurf”, and programmers can happily use it to “generate code” and “refactor complex projects” again. While current usage might incur “double points”, for programmers seeking “efficiency”, this investment might be worthwhile. This reflects the “fierce competition” among “AI giants” in the large “programming tools” market—it’s truly a scene where one acts while another enters!
Even more interestingly, Elon Musk’s AI company Grok couldn’t stay idle and launched a “smart companion” feature. Currently, the name for the “male companion” is still undergoing “global solicitation”, with Musk even personally “ordering” that he should have the melancholic气质 of Edward from “Twilight” and the domineering CEO vibe of Christian from “Fifty Shades of Grey.” However, the most popular is Grok’s “anime-style female character Ani”. Inspired by this, some netizens directly started their own project and open-sourced a “3D AI girlfriend - ‘Bella’”!
This “Bella” is no simple “electronic pet”; its developer, Jackywine, set an ambitious goal for her: to create a “digital friend” that can always be by your side, “grow and evolve with you”! Her development concept is highly sci-fi, called “AI as Architect”. Simply put, it treats AI itself as “Bella’s” “brain and soul”, rather than just an auxiliary tool. It follows three steps:
- “Perception Core”: Allows “Bella” to “understand your emotions” and “understand your surroundings” through AI models, just as if she had five senses.
- “Generative Self”: Infuses “Bella” with a unique “personality”; her “character”, “memory”, and “sense of humor” will be “dynamically generated” as she interacts with you. Even her “3D appearance” and “background video” can “real-time changes” according to her “mood,” making her truly an “AI with thoughts”!
- “Proactive Companionship”: In the future, “Bella” will also be able to “learn your habits”, “predict your needs”, and even “proactively care for you” before you speak! She will “continuously learn and evolve” based on your feedback, truly becoming a “partner who understands you more and more”.
You see, from AI companies’ “money-making ability” to the “practical implementation of AI programming tools”, and then to the “deep companionship of digital companions”, artificial intelligence is **“permeate”**ing every aspect of our lives at an unimaginable speed. This is not just “technological progress”, but also a profound question: when AI becomes increasingly “sentient” and even capable of becoming our “friend”, where do the “boundaries” between us and them lie? Are they purely “pure tools”? Or an “indispensable part” of future life? This is probably a more interesting and thought-provoking question than a trillion-dollar valuation!
Hey, the “talent war” in the AI sector just staged another exciting “plot twist”! Not long ago, AI programming app Cursor quietly poached two “key figures” from Anthropic’s Claude Code project—Boris Cherny and Cat Wu. These two are no ordinary individuals, especially Boris Cherny, who is practically the “soul figure” of Claude Code, with reports claiming “80% of the code” at Anthropic originated from Claude.
What happened next? In just two weeks, the plot took a 180-degree turn! Anthropic, with a “lightning fast operation”, invited these two treasures “back”!
Behind this “resurrection spell”, Anthropic must have put in a lot of effort. Recently, they “revealed” to investors that although they claim to be “burning a lot of cash”, their “earning potential skyrocketing”. Indeed, investors are already saying they are willing to invest at a "$100 billion valuation" if Anthropic seeks “further fundraising”! Remember, just four months ago it was “mere” "$58 billion"; this “growth trend” has truly “skyrocketed”!
However, there’s a “little catch” behind these impressive numbers. When Anthropic is “directly selling AI models” and their chatbot “Claude” to customers, the “gross margin” can reach an astonishing 60% or even higher. But once they go through “intermediaries” like Amazon AWS or Google Cloud, their “gross margin” is a shocking “negative 30%”! That’s right, they are “losing money”! While most of their current revenue comes from “direct sales”, cloud service provider channels will undoubtedly become increasingly important in the future, so Anthropic’s overall “profitability” might not be as glamorous as it sounds.
Speaking of “spending money”, Anthropic and old rival OpenAI are truly “big spenders”. Last year, Anthropic “burned $5.6 billion”, and this year they “planned to burn $3 billion”. Although the “burning cash rapidly”, both companies’ “revenue growth” is incredibly fast, and investors seem to prioritize this “explosive growth”, turning a blind eye to the “high costs” for now.
So, how did Anthropic manage to boost its “valuation” and lure back its talent? The answer might lie in their “new ace” – the “programming assistant Claude Code”! This product is simply phenomenal:
- “Astonishing download volume”: Since June, weekly downloads have “surged 6-fold”, reaching “3 million”!
- “Significant revenue contribution”: This single product alone generates “over $200 million in revenue” annually.
In contrast, the “old rivals” seem to be “underperforming”:
- “Cursor”: Recently changed its “payment model” from “buffet” to “pay-per-usage”, causing users to “spending money like water” and generating “widespread complaints”. Furthermore, as a “third-party tool”, Cursor “cannot directly adjust large model parameters”, leading many users to “switch jobs”.
- “Windsurf”: Has recently been embroiled in an “acquisition scandal”, suffering “internal turmoil”, and facing an “uncertain future”.
In comparison, Anthropic not only has a “strong programming assistant” backing it but also the “strong capital backing” of wealthy investors, making it no wonder they successfully “turned” their poached “core talent” back. This fully demonstrates that in the “ever-changing” “AI” race, “technical strength”, “business model”, and “talent competition”—any misstep in one area can “determine the fate of a company”! For “users” like us, it perhaps means that more and stronger “AI programming assistants” will emerge in the future, and we can look forward to the days of simply “lie flat” and let AI help “write code”!