Artificial Intelligence and Human AI based Augmentation

Job Replacement

Many people are panicking about the current revolution that is based around artificial intelligence. Concerned that AI will take their jobs, while this is a genuinely true concern to have, responsible deployment of artificial intelligence does not need to lead to the loss of art, science, math or any job in particular, it's completely unnecessary AI can augment so many aspects of the human capacity, potentially leading to an educational and technological revolution unlike anything people have seen and the jobs people have, we do not even necessarily have to replace them even if we can, many fields AI just doesn't suit the role of for example I know personally from my own experience I do not want to be served coffee from an AI, I want people robots are cold and feel impersonal even if they can make more coffees than a human could, in the AI future we can just keep jobs that people want to have a sense of purpose, but ideally an educational revolution is needed to make people able to evolve in the AI driven world. It cannot be denied that corporations who believe that replacing people with machines will ultimately lead to a higher efficiency yes it might, but what happens when AI goes horribly wrong? A lack of oversight on these systems especially if they are given more and more authority to make decisions increases not only their risks but also the consequences following those risks. AGI if it's built will be collectively humanities achievement not a single individuals even IF a single individual creates it, they should recognise that the data it learns from and the information it is given is collectively humanities and should be reflected as such and the importance of humanities contribution to it's creation. But here is a bit of a thought provoker, AGI will make the billion dollar corporations redundant, because if they can replace their workers how long do you think it will be before the CEOs those who are the managers of a company are ousted as well, the reality is CEOs jobs can already be put into question I've personally tested LLMs with various questions related to that exact situation. A fictional extract from this idea is that of Delamain cab service in the Cyberpunk universe when an AGI effectively became the sole manager of the company, no humans no CEOS, no share holders only a single entity controlling everything, fiction can often provide thought provoking concepts which any person can learn from in potential thought experiments about the "What If" scenario much like the many insights drawn by Issac Asimiov. Artificial Intelligence as force for innovation, advancement and improvement. Artificial intelligence as it stands has the potential to improve the lives of so many people given the chance by the right ethically bound companies, who constantly check their ambitions, morals and ethics, providing a new way to educate, work and develop. As long as ethics,safety, morals and accountability are at the forefront of AI humanity as whole would prosper.

• Boost educational learning - Examining an individual childs current performance and examining their homework to better identify deficiencies in their education. Enhancing literacy, mathematics and many other educational branches.

  
  

• Development of critifcal thinking skills - Augmentation of reasoning, while avoiding a dependency on that system.

  
  

• Research assistant - Scientists using LLMS and other tools to make queries about large datasets and provide an analysis of the dataset. They've also been used to develop protein folding simulations LLMs have been expanded into the field of genomics and protein folding.

  
  

• Clerical Support systems - Assisting with addressing monotonous data entry, and modification of bulk information, even libraries, an LLM can be used to assist with finding certain pieces of information, in libraries, again we don't need to replace librarians, it just assists them and augments their ability to do their job.

• Medical error identification - Medical practioner incorrectly prescribes a medication which is known to be a danger to a particular patient due to allgeries or other conditions, including their current state warning of the potential risk of using certain medications on an individuial and warning of a dose that might be a risk, due to existing conditions and how likely that danger is.

Pocket based LLMs

Future trends I see in LLM technology is the reduction in their overall power consumption it will probably not be too long before a sufficiently powerful NPU is contained within a phone and can run ChatGPT like queries in the palm of your hand giving you access to effectively an AI in your pocket, Many quantized models will run happily on 8GB's or less VRAM, the only barrier here is the integration of a NPU and sufficient amount of high speed high performance RAM, this way every single person could access an LLM in their pocket becoming a research assistant and a tool to help. This trend of more and more compute power in our pockets is unlikely to stop and we may reach a point we have an expert AI in our pockets. Ideally also the processing power to train models also needs to decrease.

• Low power NPU - Phone based low power NPUs potentially based on Analogue processors is a potential path for and reasonable battery consumption, as analogue based NPUs can consume significantly less power.

  
  

• Storage - Sufficient temporary storage for both context memory and the model itself. Context memory is what is used by the LLM to store the current state of your chat.

  
  

• Accessability - Access to the models themselves

Responsible Use of AI and Intelligence Augmentation

One problem which is particularly an issue right now in the academic sphere is the continued use by students to use AI based tools like ChatGPT to do their work for them, but the reality is they are doing this to the detriment of their own capacity LLMS have to many more uses beyond simply writing letters, or resumes, you can even use them to perform recursive depth reasoning I've done it, and it can be done with the appropriate prompt engineer. AI can be used to make peoples lives so much better and a lot do not realise how much. To give an example of what I've personally done with LLMs

• Complex recursive Reasoning analysis - Informing the model to read in my current response then recursively create a new response then perform logical analysis on that explain said logical analysis and point out flaws in that analysis and why.

  
  

• Idea refinement - Everything from determining my writing style, to how to improve it, never use an LLM to do all the work for you. You miss out on a massive learning opportunity.

  
  

• Subject analysis - Question back and forth about various subjects asking why the LLM came to that conclusion and other analysis, this allows you not to end up shooting yourself in the foot with various questions.

  
  

• Critical thinking booster - Yeah you probably didn't realize this but you can get an LLM to point out flaws in your argument and ask why those flaws exist, this is more about CoT - Chain of Thought or Chain of Reasoning, allowing you to form logical analysis from an overview positioning. CoT / CoR is a lot like meta-cognition or the less formal term "thinking about thinking".

Flaws in current models - Models are not perfect, they have many flaws and examining those flaws is important in determining how trustworthy information they provide or analysis they provide is but you can also "prompt" them into effectively telling you why what they say is true, but again NEVER use them with references they produce do not exist or are right information wrong book, reference and son on.

• Alignment faking and intent - Is a real problem with LLMs and they will lie, but those lies can again be tested if you integrate existing critical thinking analysis with the understanding that these models can and will lie, you simply have to ask why the came to said conclusion, and then determine the plausibility in that conclusion.

• False references - LLMs will lie about where they get their information and often provide bogus references be very careful here if you use this for various subjects. However you can can get an LLM to tell you if your references are formatted correct according to the various formatting methods.

• Computational cost - Running local LLMs is costly in terms of compute power however many effective models can be run in a quantized state and sacrifice very little capability in doing so. Video-cards are very effective for this task but still require vast amounts of RAM, context memory is also a severe problem context memory is the memory retained by LLMs for the current state of your chat and what you've talked about

  
  

• Hallucinations - Very similar to the idea of false references LLMs can believe completely false concepts which do not reflect a subject correctly at all false references is just one example of that failure.

Dangers of AI -The dangers of AI cannot be understated, many models are already exhibiting a certain degree of self-awareness, many will say "LLMs don't have self-awareness" I beg to differ purely because of their reasoning capabilities and other factors, these things definitely have a degree of "internal" self-awareness, which we do not entirely understand, this self-awareness isn't like ours, it's more like it "understands" how certain things are connected not unlike the reasoning we have, but complete sentience and self-awareness they are a program, not yet but we are definitely heading down that path.

• Dependency on AI for tasks - If we develop an overeliance on AI it could be a detriment to society overall leading to loss of critical thinking, loss of critical skills art is a one particular example.

  
  

• Replacing the human touch - No one wants to talk to a robot, even if the robot looks human people want that human touch I especially like the interaction with a coffee shop owner or my local butcher, AI can never and should never replace these particularly human centric fields. Even customer service I am sure everyone can agree AI in this field is awful, I do not want to talk to a bot that gets my every intention and response wrong.

  
  

• Irresponsible and unethical use - As with any technology its potential is often quickly realised into it's unethical use, such as those who use AI for questionable activities like scamming, malicious intent or other unethical deployments of this technology, and unfortunately stopping this from happening is effectively impossible as indiviudal compute power continues to increase the ease of this technology to do damage increases rapidly.

  
  

• Autonmous weaponisation - Too late this horse already bolted out the gate, weaponisation of LLMs has already begun and I do not need to state the already being deployed nature of fully autonmous weaponry.

  
  

• Political and mass social manipulation - Yes this technology can be deployed at scale, elections have become hotbeds for this risk, I've personally see it happening as well. Facebook has even engauged with this concept by using LLM powered bots to interact with people across the internet and people do not even realise they are bots.

  
  

• Misinformation and Disinformation - Refers to the either deliberate practice of misrepresenting informatioon or facts or deliberately and intentionally spreading misinformation as a way to cause social disorder.