AI in 2024: Exploring the Great and Horrific Possibilities

By
Wes Roth
-
42
315

The Rise of AI in 2023 and Predictions for 2024

A Recap of Google’s Advances in AI in 2023

Google has been at the forefront of AI and machine learning technology, and in the year 2023, they have continued to make significant advances in this field. From Google Core to Google Deep Mind, the company has made massive strides in AI, machine learning, robotics, and more.

In a recent blog post, Google dropped everything they’ve done in the year 2023 in regards to AI and computing, and the list was extensive. From the release of GPT-4 and Bard to Elon Musk announcing XAI and autonomous AI agents, the year saw remarkable developments in the field of AI.

As we look back on the incredible progress made in 2023, it raises questions about what’s to come in 2024. Will we see the same face-melting advance forward, or will the pace of advancement plateau? Will we hit an AI winter, as some people predict? In this article, we’ll recap the highlights of 2023 and make predictions for what we can expect in 2024.

Open Source AI Models and Predictions for 2024

One of the most remarkable developments in 2023 was the progress in open-source AI models. The proliferation of large language models and open-source AI technologies has leveled the playing field in AI research and development. Models like Bloomberg GPT and Adapt LM have shown that open-source models can rival and even surpass proprietary models, despite the vast resources poured into their development.

Based on this trend, it’s predictable that we will see open-source models catching up to the state-of-the-art proprietary AI models in 2024. There’s a high chance that we will witness the emergence of an open-source model equivalent to GPT-4 by the end of next year. Additionally, I predict that OpenAI will release a new iteration of their AI model, possibly GPT-5 or a comparable advancement, in 2024.

The Era of AI-Generated Art, Movies, and Music

The field of AI-generated art, movies, and music has seen rapid progress in recent years. With advancements in generative AI, anyone, regardless of their resources, can create incredible visual art, movies, or music. The technology has empowered individuals to unleash their creativity and produce content that rivals that of big production houses.

In 2024, I foresee the first viral movie or episodic show being generated mostly or completely by AI. While AI-generated content is not new, it’s only a matter of time before we see a breakthrough that captures the attention of the general public. Whether it’s a movie or a TV series, created by a relatively average individual, the democratization of content creation through AI is on the horizon.

The Future of Robotics in 2024

2023 saw remarkable breakthroughs in the field of robotics, with companies like Tesla unveiling their humanoid robot, Tesla bot. These advancements have showcased the capabilities of robots to understand vague instructions, reason about tasks, and perform general tasks with great efficiency. While consumer-grade humanoid robots might not be widely available in 2024, we can anticipate their increased use in production and commercial settings.

The Impact of AI on Warfare and Cybersecurity

The use of AI in warfare and cybersecurity raises complex ethical and strategic questions. The rise of autonomous killer drones and asymmetric attacks using AI technology has highlighted the need for advancements in defensive capabilities. Balancing the offensive and defensive capabilities of AI-driven warfare will be crucial in shaping a safer and more secure future.

In 2024, we may witness technologies that shift the balance in favor of defense, making it easier, cheaper, and more effective to defend against asymmetric attacks. Advancements in AI-driven defense technologies could potentially reduce the risk and impact of warfare in the modern world.

In conclusion, the year 2023 has been a pivotal year for AI and its applications in various fields. As we look ahead to 2024, the possibilities and challenges posed by AI continue to shape our future. Whether it’s open-source AI models catching up to proprietary ones, AI-generated content going viral, or advancements in robotics and defense technology, the impact of AI in 2024 will be significant and worth monitoring closely.

RELATED ARTICLESMORE FROM AUTHOR

42 COMMENTS

  4. My guess is that once we have robots we'll move back through a phase of 3d printing that will eventually allow us to print out our robots. When open-source + unregulated self printing robots hit the scene… I think then it'll be our fork in the road. If the transition replaces human workers who are left out, there's going to be some radicalization issues.

  5. I honestly just stopped worrying about things I can’t control. Just relax and do your thing. Changes gonna come no matter how much we stress over it. Our job is adapt and thrive. A friend of mine complains about AI all the time. He’s trying to stall it as much as he can, but he doesn’t realise that it’s futile.

  7. Really love your channel.. you seem to be one of the few people who doesn’t tell us about the Ferrari in his garage 😂😂.. keep delivering the truth and honest opinion .. also you feel like someone who I would hang out and have a beer and our conversations would never end ..nor would the beers 😄

  8. I disagree that defense costs will not come down.

    For example with drone attacker example, I am sure there is a relatively inexpensive way to train a seawiz to deal with that attack vertor that is significantly cheaper than the technology that precedes it as a defensive response that would be worth the investiment over the long term.

  11. Thanks for including the weaponization of AI. That'll be the big story of 2024. It's too easy to be a violent extremist now, not just because the tech makes it easier, but also because the world we've built for people too often leaves them without better options.

  13. Great, as always.

    Defending against ai means outsmarting it. That in and of itself is a lot to think about.

    More, the ability to find quickly find new and unexpected ways to attack is very difficult to defend against, and you also have ai as an adversary once those attacks are discovered/begin.

  14. I remember when I read that news about these drones that don't need a human decision to kill and I thought to myself whoever invented it should be given credit for the greatest human stupidity at that award he would have been handed a golden smart bomb… 🏆

  15. 3:30 worthwhile overview of 2023 and current vectors ("Domain Specific Corpora Influenced …") Highly simplified, essentially conversational expert systems – like more than 40 years ago.

    12:00~ starts sequences of zoom-outs of overwhelming detail – it's been about a year since I saw this attempted when it was still dripping morphing hallucinatory. But this surpasses anything I've seen in cinema, short of Google Earth.

  17. You’d think a drone built to track and bring down another drone, by definition should be lighter, quicker and cheaper than a drone that contains enough armaments to do damage to a ship. Why don’t we have effective anti-drone drones?

  18. Honestly I don’t think anything is coming true about AI other than the rudimentary features civilians will have access to, and what the corporate & military overlords what to use it for. Outside of that, it’s nice to fantasize and talk about it all in an open setting.

  19. I don't see why we act like this whole AI thing is a foregone conclusion. There isn't an algorithm that generates intelligence. It's a series of analog signal processes and reactions that cause it to emerge from a complex web of neurons (none of which we fully understand even individually). We're still trying to do this in binary, and we have no proof of that ever working in the real world. We only have examples of living creatures doing this. A simple honeybee is more sophisticated than any AI we have created ever. Low power use, highly adaptive, learns with extremely low amounts of training data, learns in real-time while it's running, and comes pre-programmed with certain evolutionarily advantageous systems.

    The only thing we have today is just complex sorting algorithms. It's what I call "deep-autocomplete". Sure it works surprisingly well, but eventually it will plateau. It has to plateau because it directly scales with training data, and compute. Once it does, we'll come up with a new algorithm that will become popular for being better than the last. Transformers are king now, but who knows what will be in 20 years. We basically just summoned a demon and tried to force it to do our homework. I really don't see this going the way people want. It rarely does…

  20. Many fear the AI revolution, much like the industrial revolution. AI Art, still needs work. As a test, give any AI Art tool a sprite sheet (animated, 4×3 or for a real challenge, try LPC) and ask the AI to change the pants to a dress, maybe add a sitting animation …. Currently, just can't do it. For extra fun try up-scaling from 16×16 to 64×64 … the AI will have no idea on what to do. Maybe in 2024 it will?
    as for war: Drone vs Drone is my prediction for 2024… both autonomous.

  27. Movies as a messaging agent is a good take, like a marketing tool. Everything initially chases popularity, and then specializes from there with profit the primary motive. Hollywood is the perfect example of this, and it's not an accident the blockbusters are all alike… they have their coveted targeted audience to influence 🙂

  31. AI is good at some things. But working with AIs for over a year, I haven't seen ANY AI so far that has even remotely any sense of judgement or insight. Any limit it has is like an unbreakable brick wall. It may be creative to some extend, and I enjoy using it. But I came to think, that entire lack of "common sense" is what will not make it replace human jobs anytime soon. A human can hear your argument and change or adapt. I haven't seen any AI do anything but refer to "I am not allowed to xyz" no matter how illogical or factually wrong it is.

    Just today I tried – for example – to make ChatGPT use Dalle3 to paint an image of Don Quixote de la Mancha. He said, nope, copyright. I replied that Miguel de Cervantes is dead since 1616, there can't possibly be any copyright. He said ok, started, stopped. Saying the same. I asked why. He could not explain, but say, well it might resemble anyone who painted it. Ok, I said, make it so it doesn't resemble anyone. Began, stopped, same reply. I asked: Okay name me style and prompts which would be allowed. He gave me prompts. EACH ONE he began, stopped, said the same. I could not get any answer as to what exactly is the reason. And that, common human sense, judgement, the ability to adapt to logical arguments, that is what a machine maybe never will have.

  35. "So my first prediction is that we will see an opensource model that is equivalent to GPT-4".

    Help me understand why this would be a big deal. (I'm very much not saying it wouldn't be, I'm truly just trying to understand.)

    If there was such a model – what would this allow me, or allow people, to do, that I couldn't do before?

    Am I correct in assuming that it will be impossible to create a model that I can run on consumer hardware, even good consumer hardware, that can compete with GPT-4?
    If this is correct, I would have to pay per use for using that opensource model on powerful hardware, via something like openrouter or poe or whatever.

    But of course, I can already pay per use for using GPT-4 via, say, openai playground.

    What I can see as a big advantage is if there was an UNCENSORED model that it is at least as strong as GPT-4. So is the big advantage that it's possible to create uncensored, or rather, "decensored" fine-tunes of opensource models, like the dolphin versions of mixtral?
    Is the big advantage that people can ANY sort of fine-tunes, custom tailored to specific use cases?
    So, say, creating a fine-tune that would be better at creating good prose than GPT-4?

    If so – is it fair to say that publishing an opensource GPT-4 level model would be the first important step, and the second important big step would be if there were pay- per-use services that allow people to use these fine-tunes of those models?
    So, say, allow me to use an uncensored fine-tune?
    If so – how are they going to defend enabling anyone to use uncensored models against government interference?

    Where, according to you, lie the big advantages in this?
    Again, I am not doubting in any way that they exist, my goal is to get a better understanding.

  36. Honestly I think lasers are going to tip the balance way against drones, at least the cheaper drones. The system itself won't be super cheap, but if you have a squad member with a backpack that can find target and fry the optics of a drone, or a more powerful vehicle-based hood ornament that can physically destroy the whole thing, any system to try to counter that will increase the cost per drone by a huge factor.

  38. As far as AI weapons and defenses go, both will be developed but never revealed to any degree. Revealing them gives opportunity to determine and develop a counter. So, both offensive and defensive AI weapons will accelerate in innovation and development, but with each building out both sides based on internal simulations against themselves. The devices will only be revealed at time of first use, at which point they'll become obsolete as the opposing side quickly develops a countering device or methodology.

  40. i have very, very strong doubts that AI would be able to simulate really strong acting in movies in the next couple of years. the level of acting required for dramas and soap operas require an emotional and embodied human intelligence that seems to not be a priority for AI researchers to develop; right now, AI researchers seem to be more focused on things like reasoning abilities, it seems to me.