I've been thinking a lot about AI for the past couple of years (as a lot of people will have) and I wanted to collect my thoughts on what I would predict will happen going forward. I'm not an expert on a lot that I will talk about so take it with a grain of salt. Also, my main motivation for writing this out is that if what I have said comes to pass, I can say that I predicted it. If it doesn't then I'll never mention it again. This is not cited or something necessarily true but something I need to write down. I might go back and add sources later. Spelling and grammar mistakes included.
So right now, most AI implementation are offered by companies, such as OpenAI, Google etc., and are not fully open to the public. A lot of talk I've read online points to the fact that this will get worse and AI will be a service offered solely by corporations. This then leads to the idea that all data will be willingly given over to those companies in exchange with excellent AI tools.
I believe this is short sighted. A big reason why AI assistants and AI tools have not integrated so well into the lives of the users is, in my opinion, due to the fact that a user never feels like they own the AI. I think the user would want the AI to adapt fully to the needs of the operator and change its behaviour. Sure companies can offer this but it will inevitably still be bound to constraints that necessarily have to be placed on the AI for legal and corporate reasons. So I think there will be a push towards self hosting AI assistants. People will be able to easily lock down their AI to only be accessible to them. Then private information like medical data or financial records can be handed over with less risk. I see two main reasons why self hosting will become more feasible and widespread. Firstly, open source AI will most likely be as good or better than closed source AI. The reason being that most of the research into AI will allow for easier creation of capable AI. Also a lot of the data sets that are used to feed the AI are already open source, meaning you can just use those. Granted, companies still hold a vast amount of data private. However, more open source data sets will continue to be created. Especially combined with opt-in data sharing from open source companies. Also, the creation of AIs such as copilot or ChatGPT will genuinely help in the creation of new AI models as people can learn and improve AI systems using the already existing AI assistants. This can and will go so far as that AI can create AI on its own without much user input. At that point any AI created by AI will most likely be better or as good as the AI that created it and a positive feedback loop would start. At this point open source AI would surpass most needs of the user.
The other reason for self hosting AI is plainly the implementation of easier self hosting systems. Right now, hosting AI on your own is a bit of a pain. They are complicated to set up, require a lot of storage and surprisingly intense hardware. Most people understandbly prefer to use online remote hosted AI services. However, as time goes on more processes will develop to help a user install and configure AI systems, taking away a lot of the complications. I predict that actually dedicated hardware for AI will start to be designed and sold, solely for the reason of self hosting AI. likely the size of a raspberry pi or a small computer, these machines could be bought and flashed with the software, eliminating a lot of the problems. Overall, like with anything as time goes on this entire process will become easier and streamlined. Most likely, AI will advise the setup of these machines, based on the requirements of the user. So all in all, I think the stranglehold of companies on AI will wane. AI assistant will become personal and private.
Next, I want to discuss something a bit less focused. I want to talk about how I think AI will change the way we look at security and the way we use our computers. I'm not a security or AI expert so I could definitely be wrong.
As companies will learn to better use the data gathered from its customers they will start harvesting even more. Instead of merely measuring clicks, watch time or engagement of users, companies will measure everything. Scrolling across a web page will be measured for speed and whether certain things are more likely to have been looked at. Recording the positon of the cursor or even the facial expressions will be common place. Corporations will become ever better at using this information to tailor the content presented to the customer to extract as much money as possible.
To avoid this, users will be able to ask the more capable and advanced AI assistants to grab webpages from a service and present it to the user without any tracking. To do this AI must become more sofisticated and knowledgable about the practices of companies. They will then be able to scrape entire websites and strip them of everything not necessary for the experience. If companies create websites which make it harder to just scrape, AI assistants would develop methods of purely presenting themselves the webpage and recreating it for the user. The AI would merely scroll through as quickly as possible and click nothing or everything making the information content nil. Companies could naturally screen for AIs as users by understanding their behaviour. However, AIs will naturally be able to in turn mimick human interactions better and circumvent the screening. This would create an arms race that only companies could lose as users could just move to a different service. Eventually companies will no longer invest into screening for AIs as any interactions with their service is better than none.
Based on this, AI would be able to mask almost any service such as apps, programs, websites, games etc. and obfuscating the unethical data gathering of the providers. Obfuscation is the name of the game as companies are no longer able to differentiate between actual and AI generated engagement. Ads will no longer be a viable way of making a service profitable as it would not be garuanteed that it was interacted by a human. However, actually interacting, like purchasing or inputting genuine information needs to still be human and can therefore seen as verified. Companies will focus on this information instead.
As time goes on, the personal AI will wrap more and more processes on phones, pcs and tablets and maybe even the actual user interface. This will lead to a new integration of AI and computers. What the user sees and interacts with will no longer just be the OS and programs but the AI presenting the information. Users will be able to customise what they can and can't see, the layour, the colours everthing. Self hosted AIs have no interests that don't align with the users. All the way the assistant will obfuscate any information sent into the web to make the user as anonymous as they would want. All of this would happen without the user thinking about it. However, explicit agreement could be enforced for the AI. Actually that leads me to the next section.
Going of the section above I want to consider how to control the AI. In general, it's quite hard to make the AI forget (it's called machine unlearning, check it out). Additionally, the systems don't always do what they should. In the future, this will be a major focus for the industry. As personal assistants will need and return a lot of personal information of the person, they will need to be controlled even more. Systems could be implemented to require agreement from the user. So sharing information about your medical history for example would need your consent, so the AI only sees what it needs to. Also if the user doesn't want certain info to be gathered by the AI (as AI will be integrated into a lot of systems, see section above) they could just say so. All past info would need to be able to be deleteable. This would further enhance the user experience and the personalisation of the AI. This is another reason why the AI would need to be self hosted, as remote hosting couldnt garauntee that the info isn't saved.
This is the first section that isn't necessarily all that optimisitc. To me, it is inevitable that AI will replace vast amounts of jobs. We can already imagine how copilot or ChatGPT will replace information based jobs. Truck drivers will be replaced by autonomous drivers. Diagnostics will be revolutionised by expert systems and many doctors will find that they aren't quite so irreplaceable. I know that through the centuries all great shifts, such as the industrial revolution and computers created more jobs for those that were replaced. However, I have a hard time imagining how an AI will not be able to do a job better than a person in half the time. Obviously practial jobs like nurses and plumbers as well as interpersonal jobs like primary school teachers will still require humans but a lot of people don't work in those industries. Here are some stats about that. So I don't know what a vast amount of the population can do to earn money. Mass unemployment is never good for the stability of a country and politics will have to catch these people. I don't know how to handle the lack of employment. However, something along the lines of a universal basic income or a continuous redistribution of wealth will be necessary. It will not be that these people don't want to work but that there aren't any jobs that they can have. This will include trained lawyers, doctors, scientists, developers and designers. How we handle this as societies will be the big question for the forseeable future.
I've used ChatGPT multiple times now to learn about concepts which I don't understand. ChatGPT is able to explain ideas and processes minutely and tirelessly. Sure it's wrong sometimes but it is better at telling me what I want to know than any person could ever be, as it knows more about everything than would be possible for a human. This'll just improve with time as well. In the future many students will use their personal AI to ask questions, create test questions and create revision plans. As the personal assistant will have all the information the user is willing to give, it'll be able to accurately tailor the explanations and questions to the knowledge of the user and the way the user can best understand it. Illustrating everything with beatiful sketches and simulations. Vocab cards will be created on the fly, audio tracks and study material will be custom generated. I don't see AI as a problem for education, I see it as another level.
I've read a lot about how AI steals art and jobs from designers and artists alike. Plays, books etc. have been written with AI or by AI. This begs the question, will AI replace creative work as well? In my opinion, AI is amazing at creation, it is unreal how well the systems write and design now and I can't imagine how good it will be in the future. It will be hard to impossible distinguishing between human and AI art soon. However, as the novelty wears off people will likely find that they don't care about what AI created. Granted AI design will permeate society and everything will see an influence. However, I think humans are naturally interested in what humans have to say. The experience and creative retelling of experience is only meaningful if humans are involved somehow. I can likely even now create a world similar in scale to Tolkien or create art that is created as skillfully as the average artist. But it wouldn't matter, I would grow bored of it. Ability means nothing if I can't know that there was some intention behind it. So I think people will value human work even more in the future. Also, I imagine physical work, like furniture, oil paintings or sculptures, will necessarily be more even more important as the AI influence will not be felt so much.
Lastly, I just want to quickly summarise my thoughts on AI age. I think we are in the AI age already. I don't know when it began but we can feel the effects everywhere. The AI age is different from the information age, because AI transforms information into something we can actually use. The AI age will be short, however, since with the advancement of AI almost every process will be optimised and improved. This will help create much better, versatile and capable robots. With those robots and the widespread use of the same the robot age will begin.