인공 지능으로 마침내 동물과 대화할 수 있게 될 것입니다.
Artificial Intelligence Could Finally Let Us Talk with Animals
https://www.scientificamerican.com/article/artificial-intelligence-could-finally-let-us-talk-with-animals/

https://www.earthspecies.org
https://twitter.com/earthspecies

제프 라스킨의 아들이자 트리스탄 해리스(소셜 딜레마의 그 분)와 협업을 많이 했던 Aza Raskin( https://twitter.com/aza )이 코파운더인 Earthspecies의 프로젝트군요~

Could we talk with our dogs and cats?

Hardware startup in UK

New computer, new UX.

https://youtu.be/MVYrJJNdrEg?si=BJ825A6d1U9nTL2t

Spatial Computing and the Metaverse: The Next Frontier in Democratizing Technology

In a world captivated by rapid technological advances, recent events like Meta's Connect Conference(https://lnkd.in/guP2dswt) and Lex Fridman's in-depth Metaverse interview(https://lnkd.in/gs4XSPYz) with Mark Zuckerberg offer a glimpse into an extraordinary future. These conversations, where real-world and digital interactions converge, hint that spatial computing could become as transformative as the personal computer itself. If made accessible and affordable, mixed reality has the potential to become the next big thing, fundamentally altering how we communicate, work, and play.

The Allure of Democratization

Just as YouTube and TikTok democratized content creation, enabling anyone with a smartphone to capture global attention, spatial computing holds the promise of democratizing our digital experiences. From Minecraft and Roblox empowering users as game developers to the vibrant ecosystems on social platforms, democratization is the wind beneath technology's wings.

The Significance of the Metaverse

The compelling interviews and demonstrations at Meta's recent Connect Conference have set the stage for what the Metaverse could truly offer. Imagine not just chatting with friends online but interacting with them as if you were face-to-face. While there's work to be done, the merging of physical and digital worlds has profound implications, from professional collaboration to social connection.

A Word of Caution

However, it's wise to heed the cautionary insights of tech veterans like John Carmack, who questions whether mixed reality(https://lnkd.in/gQ9Cde2z), as it stands, has a "killer app" to catalyze mass adoption. His skepticism serves as a reminder that successful technologies need to offer tangible utility, not just wow factor.

Lessons from the Past

The successes and failures of previous technological shifts offer guidance. The internet revolutionized communication and information access because it was both accessible and useful. On the flip side, 3D printing, despite its revolutionary potential, hit roadblocks like high costs and a steep learning curve.

The Path Forward

To make spatial computing and the Metaverse mainstream, we must focus on accessibility and real-world utility. These elements are vital in cultivating a robust user community, acting as a catalyst for wider adoption.

In conclusion, as we stand at the threshold of a new digital era, balancing aspiration with practicality becomes increasingly crucial. Informed by the past, and inspired by the likes of Meta's vision, we can aim to create a future that is not only breathtakingly innovative but also inclusively democratized.

https://www.linkedin.com/posts/activity-7113426158074957824-3zon

With many 🧩 dropping recently, a more complete picture is emerging of LLMs not as a chatbot, but the kernel process of a new Operating System. E.g. today it orchestrates:

- Input & Output across modalities (text, audio, vision)
- Code interpreter, ability to write & run programs
- Browser / internet access
- Embeddings database for files and internal memory storage & retrieval

A lot of computing concepts carry over. Currently we have single-threaded execution running at ~10Hz (tok/s) and enjoy looking at the assembly-level execution traces stream by. Concepts from computer security carry over, with attacks, defenses and emerging vulnerabilities.

I also like the nearest neighbor analogy of "Operating System" because the industry is starting to shape up similar:
Windows, OS X, and Linux <-> GPT, PaLM, Claude, and Llama/Mistral(?:)).
An OS comes with default apps but has an app store.
Most apps can be adapted to multiple platforms.

TLDR looking at LLMs as chatbots is the same as looking at early computers as calculators. We're seeing an emergence of a whole new computing paradigm, and it is very early.

https://x.com/karpathy/status/1707437820045062561?s=46&t=h5Byg6Wosg8MJb4pbPSDow

Meta starts open-sourcing a lot and is now becoming one of the best companies in the world at shipping AI features. Coincidence? I don’t think so.

Contrary to popular belief, a company (or a country) sharing their research, models and datasets publicly in open-source makes them MORE competitive, not LESS, even more so in AI. IMO, that’s how the US and some companies like Google & OAI established their leadership in the past few years IMO (even though they are not so open anymore).

Some of the reasons why open-sourcing makes companies more competitive:
- Open science and open source attracts and motivates the best talents who want to to contribute to the field
- It focuses organization on the speed of building - not on taking advantage of the current tech - especially important on a fast moving domain like AI
- It motivates the whole field to improve what you’re building on (bug fixing, optimization, new capabilities) that you can then really easily integrate in your products).

Is your company sharing their research, models and datasets? If not, they’re missing out!

Source: https://lnkd.in/e5cE93Tp

김한재님 페북. 왜 중국은 거대 모델 Race에서 뒤떨어져있나?_공산당, Tech Giants들의 복잡한 이해관계가 넘쳐나는 인재/자본

Image / Video 관련 AI Application 및 관련 분야는 솔직히 중국이 전 세계에서 가장 앞서있지 않나 싶다
당장 Tiktok만 보더라도.. AI 필터를 보더라도 그렇고. 중국에는 이미 가상의 아바타로 필터 씌워서 24시간 비디오 커머스하는 서비스도 존재한다 (!!). 이게 AI의 끝판왕이 아니라면, 도대체 무엇이 AI의 실제 application layer란 말인가.
멀리갈 것 없이.. 지금 당장 최근 CVPR paper만 봐도.. 모두 중국/중국인 연구가 압도적.
이미지/비디오 AI 가 중국에서는 처음에야 surveilance 로 시작한 태동한 분야겠지만.. 모바일/embedded/실제 사용가능한 application 분야, 아니 그냥 모든 분야에서 전체적으로 앞서있다.
멀티모달이 앞으로 대세가 될 것이란 것은 너무나도 자명하고, AI가속기 및 하드웨어 역시 이를 잘 서포트 하는 것들이 주류가 될텐데..
중국의 AI관련 HW/SW 생태계는 상당히 무서울정도의 수준과 더불어, 제재와 상관없이 어느정도 독자적인 수준에 이르지 않았나.. 라고 혼자서 지난 몇 달간 생각해왔는데..
물론, 미국을 위시한 서방권 (그래봤자 미국/영국 말고 뭐 있나 싶긴 하다. 유럽은 아무것도 못하고 있는 중) 이 앞서있는 것은 사실이지만, 중국이 AI 하드웨어/반도체 관련 제재를 받아서 스스로 나자빠질것이란 생각 자체는 상당히 나이브한 동시에, 중국의 AI 기술 성숙도 및 실제 applicable 한 제품 및 서비스가 어디까지 왔는지 잘 모를때만 얘기할 수 있다고 생각.
p.s. 중국의 cambricon 같은 애들이 뭐 만드는 애들인지 한국에 계신 분들은 다들 관심도 없으신것 같더라고. 일본의 preferred networks는 개인적으로 한껏 기대했는데.. 결국 일본애들이 일본하는거 아닌가 싶다.

https://twitter.com/sharpchinapod

New pod cast by Strachery

Graph Neural Prompting with LLMs
Proposes a plug-and-play method to assist pre-trained LLMs in learning beneficial knowledge from knowledge graphs (KGs).
Includes various designs, including a standard graph neural network encoder, a cross-modality pooling module, a domain projector, and a self-supervised link prediction objective.
It looks like a really effective way to learn and capture valuable knowledge from KGs for pre-trained LLMs to enhance them on tasks like commonsense and biomedical reasoning.
Graph Neural Prompting can improve the performance by +13.5% when the LLM is frozen, and +1.8% when the LLM is tuned.
KGs and GNNs are underrated but they are quite effective for problems where you are dealing with factual knowledge and complex structural information.
The innovative plug-and-play method significantly enriches LLMs with Knowledge Graphs. It adeptly integrates varied modules, showing marked improvements in nuanced tasks and addressing challenges with factual and structural info, making this paper key for those seeking advancements in sophisticated #AI understanding.

https://arxiv.org/abs/2309.15427?fbclid=IwAR3amz-UXFTS2_C1nCnpxUzAawbFOI2ORVxUqfTE4AKR6x1wZg48tViJy88

Here are @eladgil’s 3 tips for people building AI agent companies:

1. Build for a specific problem . Whenever there are these big technology waves, everybody tries to build things that are very general purpose. And it’s actually very useful to do the opposite and to ask, “What is the singular use case that my agent will solve?” I don’t need to develop a general purpose agent for everything. I need to solve 1 or 2 use cases extremely deeply.

2. Ship fast . Fast speed of iteration matters a lot because it’s a very competitive market. Everybody is doing a land grab, and so speed is really important. Often people wait for something to be too good before they launch it.

3. Focus on your users, not the competition. People sometimes get very competitor centric or they try to copy things competitors are doing, or they see somebody raise a giant round or whatever. It usually doesn’t matter. Just remain focused on your users.

“It’s the early days of what I think one of the most exciting moments in time in technology, at least that I’ve lived through.” - Elad

From the @agihouse_org
Autonomous Agents hackathon back in July.
https://twitter.com/i/status/1706409419129627058

VR보다 스마트 글래스의 보급이 더 빠르려나요? 스마트 폰을 쓰는것보다 더 편한 use-case를 찾을지 기대되네요.

https://about.fb.com/news/2023/09/new-ray-ban-meta-smart-glasses/

The most interesting thing about this isn’t any of those specs. It’s that these are the first smart glasses that are built and shipping with Meta AI in them. Starting in the US you’re going to get a state-of-the-art AI that you can interact with hands-free wherever you go…
This is just the beginning, because this is just audio. It’s basically just text. Starting next year, we’re going to be issuing a free software update to the glasses that makes them multi-modal. So the glasses are going to be able to understand what you’re looking at when you ask them questions. So if you want to know what the building is that you’re standing in front of, or if you want to translate a sign that’s in front of you to know what it’s saying, or if you need help fixing this sad leaky faucet, you can just talk to Meta AI and look at it and it will walk you through it step-by-step how to do it.
I think that smart glasses are going to be an important platform for the future, not only because they’re the natural way to put holograms in the world, so we can put digital objects in our physical space, but also — if you think about it, smart glasses are the ideal form factor for you to let an AI assistant see what you’re seeing and hear what you’re hearing.

I like his essay and thought framework.

https://stratechery.com/2023/ai-hardware-and-virtual-reality/

Each of these three categories, though, is distinct in the experience they provide:

Media is a recording or publication that enables a shift in time between production and consumption.
Telecoms enables a shift in place when it comes to communication.
Technology, which generally means software, enables interactivity at scale.
Another way to think about these categories is that if reality is the time and place in which one currently exists, each provides a form of virtual reality:

Media consumption entails consuming content that was created at a different time.
Communication entails talking to someone who is in a different place.
Software entails manipulating bits on a computer in a manner that doesn’t actually change anything about your physical space, just the virtual one.
The constraint on each of these is the same: human time and attention. Media needs to be created, software needs to be manipulated, and communication depends on there being someone to communicate with. That human constraint, by extension, is perhaps why we don’t actually call media, communication, or software “virtual reality”, despite the defiance of reality I noted above. No matter how profound the changes wrought by digitization, the human component remains.

I wonder what my reaction would have been to this announcement had I not experienced the new OpenAI features above, because I basically just made the case for smart glasses: there is a step-change in usability when the human constraint is removed, which is to say that ChatGPT’s vision capabilities seem less useful to me because it takes effort to invoke and interact with it, which is to further say I agree with Zuckerberg that smart glasses are an ideal form factor for this sort of capability.

However, it seems possible that AI — to Zuckerberg’s surprise — may save the day. This smart glasses announcement is — more than the Quest 3 — evidence that Meta’s bet on hardware might pay off. AI is truly something new and revolutionary and capable of being something more than just a homework aid, but I don’t think the existing interfaces are the right ones. Talking to ChatGPT is better than typing, but I still have to launch the app and set the mode; vision is an amazing capability, but it requires even more intent and friction to invoke. I could see a scenario where Meta’s AI is inferior technically to OpenAI, but more useful simply because it comes in a better form factor.

This is why I wasn’t surprised by this week’s final piece of AI news, first reported by The Information:

Jony Ive, the renowned designer of the iPhone, and OpenAI CEO Sam Altman have been discussing building a new AI hardware device, according to two people familiar with the conversations. SoftBank CEO and investor Masayoshi Son has talked to both about the idea, according to one of these people, but it is unclear if he will remain involved.

There are obviously many steps before a potential hardware product, including actually agreeing to build one. And there is, of course, the fact that Apple and Google already make devices everyone carries, with the latter in particular investing heavily in its own AI capabilities; betting on the hardware in market winning the hardware opportunity in AI is the safest bet.

That may not be a reason for either OpenAI or Meta to abandon their efforts, though: waging a hardware battle against Google and Apple would be difficult, but it might be even worse to be “just an app” if the full realization of AI’s capabilities depend on fully removing human friction from the process.

Hardware does matter — that has been the focus of this Article — but it matters as a means to an end, to enable an interactive experience without the constraints of human capacity or the friction of actual reality.

What Jobs talked about new type of computing.

https://youtu.be/T0dCm4RB63U

https://youtu.be/Q_TampOBKLM?si=XLe-DAOlWxWmWk0n