Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Muse – Meta’s personal AI agent

ai.meta.com

Post reply on HN
Print view View original

431–440 of 718 posts

Re: Muse – Meta’s personal AI agent

#431

Post by smelendez »

tyre wrote: post #398
ajkjk wrote: post #240

Earlier quoted context omitted.

I think you're right, but I'd add that... if this wasn't obvious to you without being told, you should really question your worldview a bit. A lot of people think that chat.com is an AI called "chat". Most people don't know acronym "LLM". Most people don't know about the existence of companies called "OpenAI" or "Anthropic", or that they are worth a lot of money. Etc.

TIL openai owns chat.com I hung out with my parents this weekend and heard them talking about "chat", as a noun. I assumed they were talking about ChatGPT; there being a literal chat.com makes sense why they'd call it that.

I’ve heard this recently too.

It’s confusing because of the streamer convention of referring to their actively chatting audience as “chat,” which had also entered broader slang use.

I saw one online use that seemed ambiguous, but I realized it referred to ChatGPT and that the poster was the exact type of guy who would be totally unaware of streamer culture despite being my age.

Re: Muse – Meta’s personal AI agent

#432

Post by fny »

tyre wrote: post #398
ajkjk wrote: post #240

Earlier quoted context omitted.

I think you're right, but I'd add that... if this wasn't obvious to you without being told, you should really question your worldview a bit. A lot of people think that chat.com is an AI called "chat". Most people don't know acronym "LLM". Most people don't know about the existence of companies called "OpenAI" or "Anthropic", or that they are worth a lot of money. Etc.

TIL openai owns chat.com I hung out with my parents this weekend and heard them talking about "chat", as a noun. I assumed they were talking about ChatGPT; there being a literal chat.com makes sense why they'd call it that.

This has been a thing well before chat.com even among engineers at our company.

Chat Gee Pee Tee is a mouthful.

Re: Muse – Meta’s personal AI agent

#433

Post by ihaveajob »

dd8601fn wrote: post #317
ajkjk wrote: post #240

Earlier quoted context omitted.

I think you're right, but I'd add that... if this wasn't obvious to you without being told, you should really question your worldview a bit. A lot of people think that chat.com is an AI called "chat". Most people don't know acronym "LLM". Most people don't know about the existence of companies called "OpenAI" or "Anthropic", or that they are worth a lot of money. Etc.

The joke of it is that I see those same people (ones who haven’t heard “LLM” before, don’t know what a model is or even the vaguest idea how these work, etc) have very strong opinions on the state of AI, where it’s headed, and are willing to make projections with timelines. Bill Maher is one example.

In general you can't take anyone whose job is to opine too seriously. When you're not familiar with what they're talking about, they make it sound reasonable and convincing. It's their skill after all. But now and then they talk about some topic you understand and care deeply about, and realize they're just repeating the talking points some staffer prepared for them shortly before. Most recently it happened to me with John Oliver talking about housing, reading from the standard left-NIMBY bingo card. I like him as a comedian, still.

Re: Muse – Meta’s personal AI agent

#434

Post by aryehof »

Perhaps a better overview is from Reuters …

https://www.reuters.com/business/meta-launches-ai-agent-that...

From the article:

- … rolled out … despite internal concerns that the technology mismanages its access to sensitive personal data

- The product will be available only in the U.S.initially

- A basic version of the agent will be available for free, while Meta will offer subscriptions priced at $20 a month and $100 a month for heavier usage

- People can opt out of their interactions being used to train Meta's AI models.

- Others flagged serious security flaws, like an agent routing around guardrails …

Re: Muse – Meta’s personal AI agent

#435

Post by zargon »

jddkj wrote: post #380
asdfman123 wrote: post #319

Earlier quoted context omitted.

No, after seeing how powerful AI is at my job, I definitely want AI to help plan dates, book flights, generate spreadsheets, do all the things. The only problem is, like someone else pointed out, that I don't trust Meta.

This is where I'm at as well. I've been using AI increasingly more and more at work and it has been genuinely useful. I want a Siri-like assistant that is actually useful . Muse seems like it's that tool. Except, it's Meta.

> I've been using AI increasingly more and more

> genuinely

We can tell.

Re: Muse – Meta’s personal AI agent

#436

Post by derektank »

rr808 wrote: post #284
abixb wrote: post #247

Earlier quoted context omitted.

> you should really question your worldview a bit. Yeah, I still struggle with it. I have a diverse friend circle and sometimes I forget that most of them don't have the same information diet that I do (or don't even care about tech the way I do). Maybe I should be more like them and just chill? I don't know.

Tech is a rather minor part of the world of knowledge. How many languages do you know? Do you know the styles of Renaissance painters? Slept with super-model? Ever seen an elephant stampede? Meditated by the Ganges? Read thousands of stories? Climbed a mountain so high you need Oxygen? Tech stuff is so boring.

A large number of those are far less accessible than tech, requiring either extensive travel or superlative physical capabilities (and the one’s that equally accessible don’t strike me as particularly more exciting than tech.)

Re: Muse – Meta’s personal AI agent

#438

Post by bblb »

nzoschke wrote: post #31

Why would you share data with Meta like this? The Agent Computer is super powerful, but by definition its easier than ever to build one yourself, and a the result is a personal and private computer. I know most "normal" people wont want to build their own this but there are still a lot of options that don't involve giving Meta or Grok all your data. I put some notes on all this here: https://housecat.com/blog/agent-c…

After the release of Qwen 3.8 I've been dabbling with my "Agent Computer". That is, a secondary GPU on my home computer. Intel B70 with enough VRAM to run Q6 and some context on llama.cpp, plus a ComfyUI instance. Total yolo mode with full access to my PC, --tools=all and sudo NOPASSWD.

It can do file organizing and wrangling when needed, it has simple web access with the curl tool, it can analyze media file contents. It's analyzed my local data and written several little tools (cli, TUI, GUI) to manage it automatically or give comprehensive reports. I can create new media through ComfyUI and Krita AI.

Sure, it can't book a flight or organize the family calendar for the next month, but I don't have any need for those.

edit: llama.cpp LLM model and ComfyUI image/video models can't live in the VRAM at the same time, but I let it code a little KDE widget with two buttons: one unloads the llama.cpp model and another one unloads the ComfyUI models.

Re: Muse – Meta’s personal AI agent

#439

Post by Rodmine »

artdigital wrote: post #310
theplumber wrote: post #226

Earlier quoted context omitted.

I mean nobody really uses Gemini because is just too weak…

More because Spark is not good. Gemini Flash 3.8 is an awesome model! I use it all the time Spark asks you for every single tool call, often forgets earlier parts of the conversation, is unable to use apps unless you explicitly tag them and and and Google has great models, but their harnesses and products are just bad

Google definitely does not have good models and if it does, Gemini Flash series models are definitely not them. I use the Google Gemini app daily (for searches, research etc.), everyday, for some query, gemini flash 3.8 hallucinates, does not understand context etc. in quite short conversations. I switch back to 3.1 pro and it works fine.

Which brings me to why do they get good benchmark scores? Have never really understood this. They must be optimizing just for this.

Post reply on HN