As we ease or stumble into 2024 voice tech and digital assistants are the new “currency” and likely the next big thing in AI. In 2024, AI will likely mirror (sort of) another messy international geopolitical year with light bursts of micro-progress as we wait for Gandhi in the Ukraine and Palestine. The ChatGPT4 Prometheus moment (LLMs, ChatGPT, Grok, Pi, Claude) was amazing even if the tech is kind of dumb and unreliable for legal, historical and geographical research.
The word on the streets of AI commerce is that by 2025 you won’t need your hands to do anything on a smartphone (or laptop). I’m pretty sure that we will get closer to hands-free anything and everything with new voice-tech ANN algos in the next year or two. The Chinese went largely hands-free back in 2016 or 2017 and that was a long time ago! It’s really about next-level specific NLP programs for directing an AI to do what you do with your fingers: type, edit, click, double click . . .
But will we have a really smart, even intuitive, personalized AI assistant in our pockets by 2025? Probably not!
“What will the next-level personal AI assistant feel like?”, you ask? Nothing less than a personalized version of your individuality and character in a program that learns from you to do the things you need, want or love to do. Full stop! And all the data the “app” runs on will be secure and very very private (it’s digital you after all!).
Previous software “explosions” are a good way of understanding AI software going forward. If we look back to the first consumer desktop software program explosion circa 1984 (sic; graphic interfaces + mice for text processing, drawing, accounting, databases) we see all kinds of new software programs AND NONE OF THEM solved general problems. They all solved specific things that humans did on paper. The 2008-2012 smartphone app eruptions were the same. New ways of doing old things: better process (Amazon), better perspective and creativity comprehension (Instagram), better security (for banking, email, and smartphones in general). AI going forward will be no different in the broad strokes . . .
Therefore, forget revolutionary AI assistants for 2024 or 2025. It’s not gonna happen if Elon Musk or Sam Altman don’t or can’t (more likely) pull it off. It will take a while for the next AI-genius human being (an HI) to present humanity with a new AI assistant neural network that is smaller, smarter and trained in new ways.
Feeding first-gen LLMs everything on the internet was a design error from the get-go. OpenAI should have foreseen the complexity of what they were doing before they did it and therefore had something on top of or above the LLM to more effectively and more “intelligently” deal with all that information and all those parameters. Which was Ilya Sutskever’s way to calm people down right after the OpenAI meltdown and reversal. He proposed that ChatGPT4 or 5 output would be safer if ChatGPT 2 or 3 decided on the final-final output of the system. Sort of like having an ape with a low IQ but a big brain answer a prompt that is then checked by a poodle . . .
LLMs were not programmed to understand language but to manipulate any and all language to make (up) all forms of language and images and audio . . . LLM chatbots don’t think, at least not very much.
The next AI chatbots and personalized AI chatbots will be smarter in general ways and more compact. In fact, Gemini (Google’s most recent challenge to ChatGPT4) was designed, in part, to handle various “problems” and to produce “various answers.” That’s right, more general but still a long shot from basic Artificial General Intelligence.
Given the historical progression of software applications or apps from desktops to laptops, to smartphones and then back to laptops (for must-haves like WhatsApp or Spotify), 2024 and 2025 will be similar if not identical. There will be many small breakthroughs: better digital intelligence, better design, more intuitive AI action and decision making, etc.
AI software made by the next generation of AI software “people” already happened in the fall of 2023. Humane AI kind may have bombed but it really kicked off what will come next: in-your-pocket (or on your jacket) intelligence digital tools that you can speak with and command by voice like in a sci-fi film. And this will be the next massive productivity boost of the last 550 years! No more keyboards! Your voice is both the keyboard and the mouse! Dig it!!!???
In mid-late December 2023, OpenAI leaked that they are tinkering, alongside ex-Apple star designers, with a new intelligent device that will be next-level. The rhetoric is still the same. Remember, Alexa and Google Assistant were supposed to be indispensable; but they weren’t, and neither was ChatGPT4; amazing but not quite there yet . . . like flying cars). Open AI’s new device will likely be very personalized (and will hopefully include new “meds” to counter hallucinations!!!). I think | feel | intuit it will be a device built to leverage the only thing they have: ChatGPT 5 and 6 and 7 (yawn).
And that’s my point, in the software evolutionary timeline most big breakthroughs are built upon more than a few minor or major other breakthroughs (GPUs, the cloud, better memory chips, much faster cellular and internet speeds, etc.). What comes next is built on what came before what’s next. Even mega-breakthroughs! [This was also true of mRNA drug tech. Lot’s of related breakthroughs by other experts in other fields led to the Pfizer | Moderna “moment.”] And moon-shot projects are now similar to really big projects: really diverse domain expert teams. Neuroscience, neurolinguistics, GPU designers, anthropologists, psychologists, biologists, circuit designers, and quantum computing explorers and others are the team members on the best AI teams (and here Google really stands out; they understand AI best in my opinion because they are not focused on consumer tech; GV focuses on life sciences and health and medical tech!).
The first really amazing Artificial General Intelligence systems will be built on the next “AI assistance apps” and everything else the AI universe produces going forward. Afterall, LLM chatbot are not intelligent at all; bit of a snooze really; exceptions: doc summarization and better communication skills if the user chooses a better style or wording; more elegant; less offensive; more effective writing for individuals who are not natural writers or communicators.
Software is evolutionary and the highest evolved AI-morphing-towards-AGI systems will be built to “subsume” all the AI assistance apps and all the voice tech improvements (NLP at even better levels than now; much better; much smarter!!! ; less hallucinations???) and other AI related innovations (GPUs, memory chips, multiple CPUs, etc.). And let’s not forget, Artificial Neural Networks (modeled on animal or humani or human-animal brain understandings; a domain with more question marks than check marks; still!!!) are just getting started. There are 7 or 8 types now but in 3 years there will be 20! And the new ANNs will be combos of the old ANNs with new twists in training, synthetic data, and synthetic supervision (sic).
So, in 2024-2025 there will be new things, some bigger than others. I can see a company nailing voice commands for editing text efficiently and intuitively (the ex-Apple team working at OpenAI will make their new device to do just than and more). Humane AI touted the advantages of answering an email (that has been read to you by your AI) and replying just with voice (not sure if they had a “read that back to me” function but that will be standard very soon). Remember, the AI will understand what a draft is, and it will understand “Send” . . . All that voice editing vocabulary and ease will become the keyboard of the future: i.e., no keyboards at all if you don’t like typing or pecking . . .
But no matter what happens, the next changes in AI will be incremental and cumulative and thus evolutionary. Fire never went away! And thank you Prometheus! AI is here to stay, and the AGI moment is getting closer every decade (or so . . .). In between, vast amounts of money will be ventured and made via mostly acquisitions. AI monopolies are real and there are only 5-6 on the planet. No one has more cash than Microsoft, Google and Apple. No one! Except governments! And they aren’t very good at investing are they?
Clearly AI govtech is also a killer market for vast improvements enabled by augmented intelligence tools of all sorts. I am working on a few at NexussPlus Inc. and have been since 2020! My interests lie in hybrid AI-IA-HI decision-making systems for organizations of all kinds. I feel companies will “profit” from these ideas, fair enough. But these kinds of ideas can really move the needle for humanity when democracy actually means my vote and yours count for something better: faster, fairer, easier, smarter. Tools first and then AGI system built on the tools that preceded them is my bet on the AI future . . . Feel free to email me your thoughts : just google search Ian Martin Ropke. Thanks for reading my mind!
Leave a Reply