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

281–290 of 711 posts

Re: Muse – Meta’s personal AI agent

#281

Post by theptip »

Avicebron wrote: post #244
simonw wrote: post #229

Meta AI's David Singleton has a thread about their security prompt injection defenses here: https://twitter.com/dps/status/2097404102468899131 > One threat we’re particularly focused on is prompt injection, and we handle it in layers. The model is trained to recognize and resist it. The harness marks anything coming from an untrusted source. Deterministic code checks the result. And an ensemble of classifiers runs wh…

I like that we are talking about prompt injections from you to your personal assistant. Obviously I know what they are for, but still, it's so funny that we've just normalized building software adversarial to the consumers supposedly engaging with the product. "I'm sorry simon, your request for purchasing milk this week doesn't correspond with Zuck's milk positions in the market. I've rescheduled that for next week."

You are pasting stuff from the open internet into this agent, and it’s crawling the web for you. This is the worst king of hazmat for LLMs, in one of the most adversarially challenging roles (unattended personal agent).

Just to be super clear. If you ask the agent to recommend a soda and buy it for you, and it goes to Reddit, it is going to be exposed to prompt hijacking attempts.

Re: Muse – Meta’s personal AI agent

#282

Post by redox99 »

tikotus wrote: post #259
abixb wrote: post #234

I think Meta's strategy is to capture the 'normie-tier' of AI users. I know most of us here on HN track model releases quite frequently and discuss every parameter weight out of them, but most of the world is just... oblivious? I was discussing latest in tech with an accounting friend out of curiosity, and I kept talking about tiers of GPT-5.6 (Sol vs Terra vs Luna) and asked which one she used on the desktop 'Work'…

I'm a coder and a daily driver of ChatGPT and I don't really know what Sol is or how I can access it. I follow discussions here and know of its excistence, but I can't remember its relation to other models or tiers by heart. What ever the default thing the website is giving me has been good enough for my needs for years.

Do you... code by copy pasting chatgpt web???

Re: Muse – Meta’s personal AI agent

#284

Post by rr808 »

abixb wrote: post #247
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.

> 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.

Re: Muse – Meta’s personal AI agent

#286

Post by baby »

anabis wrote: post #257

First thing I am using this is for scraping (my) groups from FaceBook since they shutdown the API a while back. "Give me comments from this group from 1/1 to 8/1 in json". Hope it works.

I use gemini all the time to check transcripts of videos! Same idea

Re: Muse – Meta’s personal AI agent

#287

Post by tikotus »

redox99 wrote: post #282
tikotus wrote: post #259

Earlier quoted context omitted.

I'm a coder and a daily driver of ChatGPT and I don't really know what Sol is or how I can access it. I follow discussions here and know of its excistence, but I can't remember its relation to other models or tiers by heart. What ever the default thing the website is giving me has been good enough for my needs for years.

Do you... code by copy pasting chatgpt web???

Yes! Though I mainly use it for one-shotting full files, so I'm not copying over small diffs. I did try copilot at some point in vscode, but after a week or so I felt like it was slowing me down. Something I knew would be a simple and quick change was now slow since I didn't have enough ownership of the code to do it quickly myself, and prompting for the change took longer than changing it manually if I had had full ownership. So I went back to coding by hand things that aren't a one-shot file.

Re: Muse – Meta’s personal AI agent

#289

Post by weitendorf »

The inline browser the agent uses that you can take control of or watch is awesome! This is something I built a lot of tools to do in the past year (including a similar inline UX/image + click pass through) but having it Just Work in a remote, fully managed client for free is extremely convenient and useful.

But… this feels like a UX that won’t last once a significant portion of consumers and purchases adopt it.

Either the network traffic is getting proxied through my client (effectively making each user a residential scraper for meta’s crawler) or it’s between meta’s servers and the sites, which puts site operators in a difficult position: if real customers are making purchases through this interface and throttling/blocking meta’s IPs makes you invisible to them and meta’s userbase, you don’t want to block that traffic.

But now every consumer in the world can just ask a question or say “check all these prices and sites for a thing I want” and go do something else, right out of the box for free, and have thousands of page loads and site interactions fire off for them.

Bypassing the branding/marketing funnels or intended UX (cf. vc twitter abusing resy thru instinct) of sites, through some kind of proxy client amplifying the traffic a human would create, with the ability to let anybody scrape or interact directly with a site’s backend… definitely a consumer win, but seems unsustainable.

Re: Muse – Meta’s personal AI agent

#290

Post by King-Aaron »

ctkhn wrote: post #267
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.

Even Shingy called chatgpt "chat" when he went on Ed Zitron's show. It was very weird to hear from a guy who has been around tech world for such a long time even if he's non technical.

I'm a pretty seasoned software dev, but I've also kept the frontier AI race at arms length from myself. I run Claude alongside my day to day workflow but don't really get overly concerned about the latest and greatest models, as I'm not leaning on it to do my job for me I suppose.

In my mind I'm after a stable tool, not wanting to dabble in the bleeding edge. I did that back in the early days of web coding and have settled into a much more subdued kind of industry/life in that regard.

Post reply on HN