Tag: LLMs

Imagining The Reality Of Our ‘Her’ Movie Future

by Shelt Garner
@sheltgarner

Here are some basic existing things that I’m using to game out the ultimate demise of the “passive Web” by, say, the end of the decade. (Excluding a pause to get the outcome of The Fourth Turning sorted out, of course.)

The first is the growing power and popularity of Large Language Models. Another is the growing power of computing hardware in general. Another is the fact that there is some techno-cultural determinism in LLMs solving some pretty basic problems with the Web as we know it. And last would be improving XR technology.

Ok, with all that in mind, it’s pretty easy to imagine that we literally could live in a Her movie-like future pretty soon. Everyone will wear a earpiece that allows them to have real-time conversations with a LLM digital assistant. The LLM’s dataset will be the entirety of the Web. The LLM will be finely tuned to your specific personality to the point that all of today’s bitching and moaning by conservatives about chatbot “bias” will seem quaint and silly.

Instead of searching Google, we will have casual, personal conversations throughout the day about whatever it is we need to know. Websites will no longer exist — or will no longer get the (human) traffic that they get today. And here is where XR technology fits in.

Instead of the passive use of a “browser” to read a New York Times article, you will actively be presented the contents you’re interested in by your LLM — probably displayed as a multimedia AR experience at your demand. The connection between AI and XR is not as obvious as maybe it should be. It seems to me that the two will feed on each other so that both are trillion dollar industries by 2030. (Again, baring the United States collapsing into civil war / revolution because of fucking ding-dong Trump.)

The thing about it is, all of this could happen really, really fast. Within just a few months, content on the Web could collapse into a Singularity with little or no direct human interaction with it. An entire genre of media will no longer be relevant at all.

Even all the microblogging services that have popped up might no longer needed because the real-time news element of the services will be replaced with you having a really interesting, personal conversation with your digital personal assistant.

Or, at the very least, what you might previously get via Twitter or Bluesky, you will see via a XR display that you wear most of the time.

Besides the obvious historical and political obstacles to this happening, there might be some human resistance to what seems to make total sense today. As such, maybe it’s not 2030 that no one reads The New York Times website anymore, but, say, 2033 or later.

But it’s coming. I just don’t see how the Web continues to exist in its present form.

The End Of The Web

by Shelt Garner
@sheltgarner

The crux of much of the money made in the Internet era has been slapping ads on solving the problem of there being a deluge of information on the Web. But it is growing more and more apparent to me that it’s at least possible that as we careen towards a “Her” movie type future, that the Web could collapse into something akin to a technological Singularity.

The entire modern Web paradigm will evaporate into into simply interacting with a hyper-personalized LLM. So, rather than a Google-style one to many situation we have now where we all go to Google to ask a question, we will each have a LLM specifically tweaked to our personal proclivities.

What’s more, the Web itself will no longer exist.

There are elements of this potential future that I can’t quite game out yet. What about video or music? How does any one make any money off of content if all content is simply fed directly into a LLM that then, in turn, tells end users about that content?

I think we have to contemplate the idea that the next trillion dollars in tech will be made from whomever can scale Her-like software. So, rather than a smart phone, you’ll have some sort of device that interfaces with a LLM. The end product will be much like what is seen in Her or even the Apple prototype commercial from way back when where a dude talks to a high end digital personal assistant.

As such, it seems Apple would be the most obvious company to benefit from this because they don’t have a profitable search business to protect AND they have a unified software and hardware ecosystem.

The issue of online content, however, is a much more difficult thing to process. But it is easy to imagine that LLM hooked up to a real-time feed of the Internet could market the final death of all media online. If your LLM uses the ENTIRE INTERNET (including the Dark Web) as its dataset, the fact that at the moment you only can get “suggestions” about a portion of the Internet’s vast content becomes quite moot and quaint.

The LLM will simply digest all of human knowledge and give you a specific answer specifically tailored to your personality and needs. It will know everything about you and so there will be something akin to “preemptive search” that takes place.

But, again, how will some future interation of The New York Times make any money? I honestly don’t know at this point. Maybe they will be paid to feed their content directly into the LLM? But I do know that the Web as we currently conceive of it is lurching towards its doom.

The Potential Implications Of Google’s Gemini LLM For Hollywood

by Shelt Garner
@sheltgarner

Apparently, Google is coming out with a next-generation LLM called Gemini later this year, in December. It’s allegedly going to be four times as powerful as OpenAI’s ChatGTP4.

The thing I’m interested in is the implications for Hollywood. It’s well within the realm of possible that the two strikes that Hollywood is experiencing at the moment will still be going on. If Gemini turns out to be as powerful as Google claims, I wonder if it’s possible that Hollywood suits might begin to turn to Gemini to make the strikes…moot.

Or, put another way, it could be that all my “hysterical doom shit” about the future of Hollywood could happen a lot sooner than we might think. It could be that the Hollywood suits will simply sit on their hands until Gemini is up and running. Then, this spring, they will start to spit out the first AI-generated movies.

Or not. I don’t know enough about what’s possible. But the point remains — things are moving very quickly with LLMs and I think we have to begin to understand that Hollywood is about to go through a massive technological shift in ways that we can barely begin to understand.

I’m At A Loss As To What To Think About Google Scraping Google Docs To Train AI

by Shelt Garner
@sheltgarner

There is nothing short of panic on Tik-Tok from writers like me who use Google Docs. A number of people have Tik-Tok have urgently suggested that all writers take all of their writing off of Google Docs immediately and use Word instead (or whatever.)

I find this very curious for a number of reasons.

I understand where these writers are coming from, and, yet, I’ve actually looked into the scraping Google is doing and it seems a little too late to worry about such things. Now only is AI everywhere now, but I just don’t know how much my writing is going to make a difference if Google is scraping the entirety of Google Docs to train their LLMs.

Some of the hysterical talk on Tik-Tok seems just that — hysterical doom shit that assumes there’s some way to prevent, completely one’s words from somehow, someway being used to train LLMs.

But I will at amit that I just don’t know. Maybe I’m missing something. Maybe if I just turn to Word I can somehow, magically, prevent anything I write from finding itself in the maw of AI.

At the moment, I’m taking a very measured wait-and-see approach. While I understand that if I keep my writing on Google Docs that it may, in some way, influence Google’s LLMs, it’s not like my, specific writing and ideas are going to magically pop up somewhere and be produced by Google Movies.

Or maybe I’m wrong. Maybe that’s exactly what is going to happen. I just don’t know at the moment.

Of AI….& Tik-Tok Reading Our Fucking Minds?

by Shelt Garner
@sheltgarner

It hasn’t happened that much recently, but Tik-Tok continues to seem to have a very…eerie…ability to pin down exact elements of my internal monologue. It’s all very curious. I’ve kind of learned to shut up about it, though, because mentioning my fears on this subject causes people to look at me funny.

But Tik-Tok continues to figure me out on such a specific manner — seemingly in real time — that there are two solutions to the issue. One, Tik-Tok has really good AI to the point that it can figure what’s going on in my mind simply from analyzing my usage…or it can fucking read not just my mind, but everyone’s mind!

I don’t think Tik-Tok is reading my mind, but I do think that however it’s figuring out what’s going on in my mind should be regulated — hopefully out of existence. What is the spookiest about what Tik-Tok does is the near-real time nature of it.

If what was going on was a general figuring out of me as a person, it wouldn’t be such a big deal. What is alarming is I think about something really hard and….then with 24 hours I get pushed some obscure thing on Tik-Tok that makes very clear reference to what I was thinking about.

Very strange. Very odd.

But absolutely no one listens to me. And if they did listen to me, they would think I had lost my mind. So, no. Like I said, I don’t think Tik-Tok can read my mind….but something…..strange…is going on.

What, though?

Is Madison Beer….Real?

by Shelt Garner
@sheltgarner

I’m not being serious, obviously, but I sometimes wonder if Madison Beer is a a Simone kind of AI-generated personality. I’ve never heard her voice, even though she’s a “singer.” She’s got a perfect smile and a great body. So, while, yes, she’s obviously, real, we’re careening towards point where you could actually pull off a Simone-type AI generated celebrity, parasocial relationships and all, pretty easily.

I say this because it’s going to happen soon enough, one way or another. Once we figure out if America is going to have a revolution / civil war or turn into an autocracy, the thing thing we’re going to address is all the amazing cultural changes that are going to happen because of AI.

We may have to bomb ourselves into oblivion to get the “cool stuff” of AI, though.

Anyway. I think all human-generated art is probably going to fade away. It’s just a matter of how long it takes. It may be 10 or 20 years, but, in the end, Hollywood as we know it — actors and all — will be more industry built on software than celluloid.

AI is going to sucker punch Hollywood actors, directors and actors in a pretty astonishing way, leaving producers sitting pretty. Will a Simone of the future be able to win an Oscar? That’s a question we’re going to have to ask ourselves as a species, I’m afraid.

The End Of (Human Generated) Art?

by Shelt Garner
@sheltgarner

As we all wake up from the Superhero movie era daze we’ve been in for 20 years, I think maybe we aren’t asking the right question. The question isn’t, “What type of movie genre will replace Superhero movies?” but, rather, “Is this the twilight of not just mass media, but human generated art?”

It’s very possible that a lot sooner than you might think, AI sensors on your TV and phone hooked up to “Her”-like technology will generate movies, TV shows, songs and — gulp — novels, that are designed specifically for what you want at that specific moment because of your mood.

Some 99% of what is generated by the hand of man when it comes to art is shit and, as such, 99% of all art could very well be, well, AI generated shit. AND, there won’t be any mass media anymore, no shared reality.

As such, the Singularity won’t come with us uploading our minds into computers, but with the very idea of human-generated art for profit being seen as a quaint notion of a bygone era. And that dystopian nightmare is probably going to happen a lot — A LOT — sooner than you think.

If it happens at all, that is.

And add the general tendency to name, shame and drug anyone who is “different” and, lulz, our posthuman future may already be here.

The WGA Is In Trouble

by Shelt Garner
@sheltgarner

As I often say — I’m always, always wrong. But it definitely seems as though it’s at least possible that we’re in something of a waiting game when it comes to the current WGA strike.

And the two sides, tragically, are waiting for different things.

The WGA is waiting for the Suits to come back to the negotiating table, while the Suits are waiting for LLMs to advance to the point that Hollywood writers are…moot. So, rather than “September” being the deadline as one very young and naive striking WGA writer proposed, I think we have a far more open-ended situation on our hands.

It could be 18 months before there’s any resolution to the Writers’ Strike and the resolution will be that technology has reached a point where the Suits feel like they can just ignore the WGA altogether. And, rather than thinking about a WGA strike, they’re thinking about how many programmers they’re going to have to pay in place of them.

Like I said — I’m always, always wrong. So, I suppose it’s possible that something will happen and the Suits will come to some sort of agreement with writers. But..I couldn’t count on it.

A Breached Second Draft Scene Birth

by Shelt Garner
@sheltgarner

I’m really struggling with a few scenes in the second draft. Two scenes in particular at the moment have pretty much caused everything to come a screeching halt.

I’m going to have to give them a lot of forethought tonight before I hopefully can turn around tomorrow morning and write them out. Not even using AI has been able to help me with these scenes because what I need from them is too complicated and, in a sense abstract.

So, I’m just going to have to go back to the old way of thinking really hard and struggling with how I can fill the scene with all the information that I need within it. Another problem is subsequent scenes really depend on what I lay out in these two scenes, so it’s difficult for me to just breeze past them and come back to them when I feel better.

Ugh.

Anyway, I continue to be impressed with how effective AI is in aiding me developing this novel. It’s definitely, in general terms, speeding up the process. But’s far from perfect.

I still have to actually write the novel out. But it definitely is filling a gap that I’ve long had because I have no friends no no one likes me. I no longer feel like I’m doing all of this in a complete creative vacuum.

Of Usenet News & LLM Datasets

by Shelt Garner
@sheltgarner

The last I checked, Google has a nearly-complete archive of Usenet from its founding until at least around 2000 when everything went to shit there because of porn spam. It would be dumb for Google not to include Usenet in any Large Language Model dataset.

You would have to tweak it some, of course, but there are about 20 years of high quality words to use to train your LLM to be found with any Usenet archive. A lot of is outdated and full of vitral, but there is also a lot of human interaction and humanity to be found there, as well.

This is so much the case that if you were to include Usenet archive information in your LLM training dataset, you would probably endup with a very human-like LLM. I don’t know, maybe Google is already using their Usenet archive. Usenet was very popular back in the day.

Given how many Usenet servers there were at one point, I’m sure if you were working on an open source LLM that you could probably find a few million words to train your open source LLM by scooping up all the archived Usenet posts you could find.

Or not, what do I know. But it is an intriguing use for all those words that are now just forgotten Internet history. For everyone except for me, of course. 🙂