Matt Invest

[Matt Invest @mattbin_gg]

@skalskip92 But is it possible to do in real-time? I tried their notebook and found out every time need to process every frame

[Gavin Baker @GavinSBaker]

1) Would be interesting to know how many of the analysts that are modeling GPU demand out 3-5 years are thinking about the impact of stuff like MCTS, other search techniques, conditional routing models, LoRA and a host of yet to be discovered innovations.

[fin @fi56622380]

终于，苹果最近发布的AR/VR职位描述里，总算泄露出来了他们要把LLM放进AR/VR设备里的想法：“Conversational and Generative AI”

从技能描述看是做app，LLM看起来更像是放在设备上，而不是云上

从使用上来看，提到了需要prompt engineering和prompt fine tuning，看起来更像是人和XR交互方式上做文章

至于更笼统的generative AI职位描述，也可能是内容生成方面做文章

从这个职位描述基本上可以验证之前提到的，在交互方式和内容产出上，LLM对AR/VR是一个很好的补充这个说法

苹果下周的WWDC大家也是对新的MR设备期盼已久，从传闻来看，第一版发布的可能是3000美元的高端开发版，之后的roadmap里再降低价格到1000美元以下和Meta竞争，这也意味着AR/VR进入大众主流市场方向迈出了一大步

-------------------------------------

AR/VR市场大概的割据格局，其实早就已经很明晰了，除了苹果和Meta之外，最大的玩家应该是Google+三星+高通合作的产品(Google之后也许会自立门户)，此外国内的一众厂商也不可忽视，pico+一众手机厂商(小米华为oppo/vivo)应该都不会缺席

格局大概可以从四个方面来看：芯片，OS，设备，生态

芯片OS方面来说，最主要的几个大玩家，分成两个派系，高通系和苹果系

苹果的芯片基本就是M2芯片魔改一下搭配IOS，这在一年前都不是秘密了，非苹果系芯片基本上是高通一统天下，骁龙芯片魔改一下，别无第二家。一个反直觉的事情是，高通这一代的GPU能耗比竟然能超过苹果GPU 20%，而等明年Nuvia CPU进入高通芯片后，在CPU上也丝毫不落下风，苹果第一次在硬件上遭遇了如此强劲的对手(当然了，iOS能扳回来太多分数)

设备上基本上就是看各家自己的底蕴了，这一块没啥好说的

这里不得不佩服一下Meta，决心真的是太强了，连年无底洞一样的投入，到今年已经到了一年160亿美元以上的烧钱速度，AR/VR虽然仍然还不算成熟，但有了LLM的加持可能会是完全不一样的局面，即便AR/VR这一波不会像iphone那样的人机交互革命，也是大大的走近了一步

算上AR/VR未来几年还需要继续烧钱的程度，Meta在这个领域的投入可能会超过一千亿美元，这已经是堪比一个登月工程的巨大投入了(当年的阿波罗计划总投入换算过来也就是这个量级)，真是佩服小札的决心

熬了这么些年总算熬到了快看到曙光的时候，虽然走了不少弯路，内部摸索阶段浪费了惊人的巨量资源，接连放弃了自研芯片又放弃了自研OS，但他们在生态上的积累，已经是毫无悬念的领先了，价格比苹果也亲民很多，这也是为什么我觉得Meta可能是最大的两个AR/VR玩家之一，因为生态的建设真的太重要了

AR/VR时代元年，一定不是2016，也许不是2023，而即便不是这两年，我们可能也比较接近了

[tom 🖍 @tomthetrader1]

S&P 500 thoughts: The market is in a bad health and I think it is mostly going unnoticed. Pay attention.

The S&P 500 closed at the highest level in 9 months (Aug 2022), and the Nasdaq at a 52 week highs. This headline reads like a bull market, but I'm not so certain it is.

Market breadth remains atrocious. More stocks are making new lows than new highs. This highly concentrated rally is uncharacteristic of rallies that end bear markets. Those rallies exhibit improving breadth.

Sentiment is fading while prices are rising. In this bear market, sentiment peaks have preceded all declines.

The AI-anything vibe on Wall Street is a hallmark of euphoria. In hindsight I think this will be very obvious. (For the record, I think AI tools are fascinating).

Debt ceiling talk, rate hikes, and the Federal Reserve have tranquilized market participants. The news cycle prior was the entire coronavirus narrative & ongoing war story. Participants are numb. (I don't blame them)

I have actively shared that history shows that if the S&P 500 can advance 5% above the long term moving average (200d exponential) it is most probable that the bottom is in and a genuine market recovery is underway. This has yet to happen after multiple attempts, and I remain doubtful that it will considering the poor breadth, fading sentiment, and mostly numb participants.

If the S&P 500 can break the 5% long term moving average threshold (currently 4245) for a week I intend on reevaluating my market outlook. Until then, I remain comfortably bearish.

$SPY $SPX

[Rihard Jarc @RihardJarc]

One of the best interviews I read.

Former Global Head at $GOOGL /GCP who worked there for 7 years:

- $GOOGL's edge on AI
- Talks about CEO of GCP: T.K - 60% of the time on customer-facing calls
- Transition from SMB clients to enterprises
- Thinks $MSFT Azure will be num. 1

[Matt Invest @mattbin_gg]

@Kumamontw @ambersweekly 已經存了 下架不了

[Matt Invest @mattbin_gg]

@mengkunghsieh 我以為server已經是已知 AMD還是跌爛 可能是蘇媽答的超爛🌚

[Gavin Baker @GavinSBaker]

1) “God-like AI. A superintelligent computer that learns and develops autonomously, that understands its environment without the need for supervision and that can transform the world around it.”

“It’s possible from now onwards.”

Worth reading.

https://t.co/XkNuGiMz9p

[Gavin Baker @GavinSBaker]

1) The strong perf of megacap tech in Q1 (up 31% per GS) is strange to me.

They were all in their own monopolistic swim lanes for 10+ yrs.

Now they are all facing an existential threat/opp from Generative AI/LLMs.

[Sebastian Raschka @rasbt]

Prefix tuning and adapters are the 2 out of the 3 most widely used parameter-efficient finetuning methods for large language models (LLMs).

If you look closely, the recent LLaMA-Adapter method that made big waves is actually a prefix tuning method, not an adapter method.

1/2

[Mahesh Sathiamoorthy @ COLM2026 @madiator]

Paper: HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face from MSR.
https://t.co/wZXssQlrvZ

I came across this paper a few weeks back and I can't get it out of my head! It's such a powerful idea that shows us how things will evolve in the future.

🧵⬇️

[Kevin Fischer @kevinafischer]

Ok this might be the coolest paper of our generation

Realistically simulated human agents, interacting together forming relationships, bonds, identities, studies through interviews…

So basically Westworld

https://t.co/tKO1xY80N4

[Matt Invest @mattbin_gg]

@cwolferesearch @memdotai mem it

[Alf @MacroAlf]

November in global macro was extremely interesting - let's dissect what happened and what might lie ahead.

A short thread.

1/

[Dylan Patel @dylan522p]

Intel earnings thread /1
Datacenter and AI group did totally horribly...
Literally no operating profit.
The worst quarter in the last couple of decades of the history of datacenter.
$INTC $AMD $TSM $NVDA $MRVL

[Gavin Baker @GavinSBaker]

1) A thread of hypotheses, loosely held.

This recession will be nothing like 2002, when EPS for the companies in the S&P 500 tech GIC declined 80%. Tech EPS likely even more stable this time than in 08/09 when EPS declined only -3%ish. Function of shift to recurring revs.

[Alf @MacroAlf]

Bond vigilantes are truly back.

The bond market already strong-armed the UK.

And if this ''pivot'' chatter is to be believed, the bond market will also strong-arm the Fed.

An important thread on bond markets and systemic risks ahead.

1/