https://x.com/tlalexander/status/1765122572067434857
These systems are making the internet a worse place to be.
Reliable information and communication will soon be opt-in only. The "open internet" will become an eclectic collection of random experiences, where any real human output will be a pleasant, rare surprise, that stirs the pot for a short blip before it is assimilated and buried under the flood.
It is under threat now, due to the increased sophistication of spam.
Similar in my current job btw.
One’s nephew might not interact in a professional capacity with other organizations, but many do.
B2B probably still uses email, but that's terrible because of the phishing threat. Professional communication should use webapps plus notification pings which can be app notifications, emails, or SMS.
People to people at different organizations again fall back to texting as an informal channel.
I have sent 23 emails already today to 19 people in 16 organizations.
Even our clients use slack to connect, it is many times better then emails.
I'm trying to think of 1 single thing email does better than slack in a corporate communication.
Is slack perfect? Absolutely not. I don't care that I can't use any client I want or any back end administration costs or hurdles. As a user, there is no comparison.
The thing I hate the most is that people expect to only use Slack for everything, even where it doesn't make sense. So they will miss notifications for Notion, Google Doc, Calendar, Github, because they don't have proper email notifications set up. The plugins for Slack are nowhere near as good as email notifications as they just get drowned out in the noise.
And with all communication taking place in Slack, remembering what was decided on with a certain issue becomes impossible because finding anything on Slack is impossible.
But no better than email again. You just get notified if you have a email or not [depending on the client you're forced to use]. I have many email rules that just filter out nonsense from github, google and others so my inbox stays clean.
I guess I find them useful because that way I get all my notifications across all channels in one spot if you set them up properly. Github and google docs also allow you to reply directly within the email, so I don't even need to open up website.
In Slack the way it was setup was global, so I got notified for any update even if it didn't concern me.
- Long form discussion where people think before responding.
- Greater ability to filter/tag/prioritize incoming messages.
- Responses are usually expected within hours, not minutes.
- Email does not have a status indicator that broadcasts whether I am sitting at my desk at that particular moment to every coworker at my company.
#2 I agree with you, again Slack is not perfect but better than email [my opinion]
#4 Slack has status updates, email does not. So you can choose to turn this off, again boundaries.
Regarding #4, I can't turn off my online/offline indicator, I can only go online or offline. I can't even set the time it takes to go idle. These are intentionally not configurable in slack. I have no desire to have a real time status indicator in the tool I use to communicate with coworkers.
There are certain actions I have to use email for, and it feels a little bit more like writing a check each year.
And all these email subscriptions, jeez, I don't want that. People who are brilliant and witty and information on Xitter or Mastodon in little slices and over time I still don't want to sit down and read a long form thing once a week.
In a battle between AI and teenage human ingenuity, I'll bet my money on the teenagers. I'd even go so far as to say they may be our only hope!
Search may be reduced to nigh uselessness, but the saavy will still be able to share quality information as needed. AI may even assist in that with people who have the domain knowledge to properly correct the prompts and rigorously proofread. How we find that quality information may, once again, be through closed off channels.
I’m not really saying that, and haven’t put a lot of thought in to my views there. But when people say the biggest problem is the controls on the AI, I feel compelled to point out that the destruction of the open internet is happening despite these controls, and is a major problem unto itself.
In other words, it's very likely generative AI will be very good at creating fake simulacra of reality, and very unlikely it will actually be good AGI. The worst possible outcome.
The older generations are no better. While ProPublica or WSJ put effort into their investigative journalism, they can’t compete with the volume of trite commentary coming out of other MSM sources.
Generative AI poses no unique threat; society’s capacity to “think once and cut twice” will remain in tact.
While the threat isn't unique the magnitude of the threat is. This is why you can't argue in court the threat of a car crash is nothing unique even when you're speeding vs driving within limit.
But a car going into a stroller at 150mph versus 200mph is negligible.
The democratization of generative AI would increase the number of bad agents, but with it would come a better awareness of their tactics; perhaps we push less strollers into the intersections known for drag racing.
I guess when you distort every argument to an absurd you can claim you're right.
> but with it would come a better awareness of their tactics
I don't follow. Are you saying new and more sophisticated ways to scam people are actually good because we have a unique chance to know how they work ?
Handwringing over the threat of AI without substantiating an argument beside “enabled volume” is just self-righteousness.
AI isn’t posed to shift the balance of MFA versus phishers in a way that can’t be meaningfully corrected in the short and long term, so using “scamming” as a means to oppose disseminating tech feels reductive at best.
It is, because I wasn't directly comparing AI to traffic but only reaching for an example to illustrate how irrelevant is the case whether the threat is something completely unique or not.
> Handwringing over the threat of AI without substantiating an argument beside “enabled volume” is just self-righteousness.
Dismissing it as "meh, not new" is plain silliness.
> AI isn’t posed to shift the balance of MFA versus phishers in a way that can’t be meaningfully corrected
What on Earth makes you think that ? The beautiful way we're handling scams right now ? If you think it's irrelevant that phishing via phone call can now or soon be fully automated and the attack may even be conducted using a copy of someone's voice - well, we won't get anywhere here.
The way we mitigate scams today definitely encourages me; the existence of victims does not imply the failure or inadequacy of safeguards keeping up with technology.
While AI stokes the imagination, it’s not so inspiring that I can make the argument in my head for you about why humanity’s better off with access to these tools being kept in the hands of corporations that repeatedly get sued for placing profits over public welfare.
Ok, now you're being just stubborn. No, no one is manually dialing your number but as soon as the scammer knows you've answered you get to talk to a human who tries to convince you to install a "safety" app for your bank or something. THAT part isn't automated, but it may as well be, which means phishing calls and scams can potentially be done with a multiplication factor of hundreds, maybe thousands - limited only by scammer infrastructure.
It could be the return of real journalism. There is a lot to rebuild in this respect, as most journalism has gone to the dogs in the last few decades. In my country all major newspapers are political pamphlets that regularly publish fake news (without the need for any AI). But one can hope, maybe the lowering of the barrier of entry to generate fake content will make people more critical of what they read, hence incentivizing the creation of actually trustworthy sources.
It's not like AI invented clickbait, though it might have mastered the art of it.
The convincing bullshit problem does not stem from AI, I'd argue it stems from the interaction between ad revenue and SEO and the weird and unexpected incentives created in when mixing those 2.
To put it differently, the problem isn't that AI will be great at writing 100 pages of bullshit you'll need to scroll through to get to the actual recipie, the problem is that there was an incentive to write those pages in there first place. Personally I don't care if a human or a robot wrote the bs, in fact I'm glad one fewer human has to waste their time doing just that. Would be great if cutting the bs was a more profitable model though.
Personally, I highly dislike this handwaving of SEO. SEO is not some sinister agenda following secret cult trying to disseminate bullshit. SEO is just... following the rules set forth by search engines, which for quite a long time is effectively singlehandedly Google.
Those "weird and unexpected incentives" are put forth by Google. If Google for whatever reason started ranking "vegetable growing articles > preparation technique articles > recipes > shops selling vegetables" we would see metaphorical explosion of home gardening in mere few years, only due to the relatively long lifecycles inherent in gardening.
SEO is gaming Google's heuristics. Google doest generate a perfect ranking according to Google human's values.
SEO gaming is much older than Google. Back when "search" was just an alphabetical listing of everyone in a printed book, we had companies calling themselves "A A Aachen" to get to the front of the book.
I fail to see immediate disagreement here: I don't see how Google's ranking process/method/algorithm being heuristic changes the observation that to a website in the end it is a set of ranking rules, that can in some ways be gamed. SEO is two part process: discovering those ranking rules and abusing them.
Your example with gaming alphabetical listings only reinforces the idea that SEO abuses rules set forth by the ranking engine.
However, it does not meaningfully matter whether the incentives inherent in the ranking system form results the way they do intentionally. What matters is the eventual behavior of the ranking system. Mostly because by definition you cannot filter out bad actors entirely, all you can do is 1) place some arbitrarily enforced barriers, which are generally prohibitively costly 2) place incentives minimizing gain of bad actors.
To clarify, Google defines the metrics by which pages are ranked in their search results, and since everyone want to be at the top of Google's search results, those metrics immediately become targets for everyone else.
It's quite clear to me that the metrics Google have introduced over the year have been meant to improve the quality of the results on their search. It's also clear to me that they have, in actual fact, had the exact opposite effect, namely that recipes are now prepended with a poorly written novella about that one time the author had a emotionally fulfilling dinner with love ones one autumn, in order to increase time spent on the page, since Google at one point quite reasonably thought that pages where visitors stay longer are of higher quality, otherwise why did visitors stay so long?
And if no one is using the search engine, who's gonna see all those ads?
Of course, they do have other incentives too, some of which are directly conflicting with high quality search results, as you point out.
I suppose one could argue that their near-monopoly in the search business has allowed them to be somewhat negligent on the quality of search, but now that there are a few competitors at least somewhat worthy of the name, one can hope high quality results will be a higher priority.
Anyway, I'm holding out hope someone will one day manage to train a LLM to distinguish between quality content and SEO bullshit, and then put that to use in a search engine.
I'm not well versed enough in the current status of LLM's to make a prediction on how hard that will be, but my impression is that we're a fair ways off from any LLM being able to do that well enough to be valuable.
I'd really love to be proven wrong on this one, if you're reading this and have some relevant experience, consider yourself challenged! (feel free to rephrase this last bit in your head to whatever motivates you the most)
But when that GAN passes the human threshold, suddenly you can use the faker to create interesting things and not just use the discriminator to reject fakes.
I can't believe this went all the way to AI...
You might as well bemoan gravity.
you have no evidence whatsoever that llama models are being used for that purpose. meanwhile, twitter is full of bots posting GPT refusals.