America On Fire

@IIIBYRDIII

Patriot to the core. I'll never be sponsored because I cuss a fuck ton. Stop giving in to the mentally ill. I piss people off. Drop a follow.

US Citizen, Prior service/LE

Joined November 2014

201 Following

276 Followers

5.1K Posts

Pinned Tweet

America On Fire

@IIIBYRDIII

10 months ago

The advancement of AI over the Years AI is a very hot topic as of today, being utilized in almost every field from farming to pharmaceuticals it's embedded in nearly every industry. Beginning in 1950 Alan Turing asked the question that led to the development of today's current version of AI and that question was “Can machines think?” That simple question developed what is now known as the Turing Test. A simplified explanation of the Turing Test is like this; You have three rooms, one room has a person asking questions in text only to the other two rooms. Within the other rooms could be a real person or a computer answering the questions. It is the person who asked the questions responsibility to identify if either room is human or computer. If the person asking the questions cannot verify which answers are human or computer then the computer has passed the Turing Test. The first documented pass of the Turing Test was in 2014 but is highly criticized by numerous experts. This will be important to note for later. The Infancy of AI When you think of AI and where it has come from you have to consider that it has been under research since the 1950s. The progression for decades was slow and lacking any crucial advancements until hardware limitations became less restrictive. In 2012 the current model of what we use today was born. No longer was AI simply attempting to solve equations, translate languages and predict the moves of an opponent in chess it now was at a breakthrough that a majority of the population use today – Speech and Image Recognition. Just 13 years ago the foundation for what is now Google Lens, Claude and Grok was laid. Nearly every AI in use today branches off from those breakthroughs in simple image and speech recognition but it still had its flaws. The training models were still in their infancy, when given information that they had been trained with numerous comparisons the AI performed above expectations but when introduced with new data that was not part of the previous training it stumbled and fell on it's face. One example from IBM's “Watson” specializing in oncology is noted to have been giving unsafe treatment recommendations to doctors eventually leading to the 62 million dollar project being canceled. It may have won in Jeopardy but it lost in its diagnostic approach. In 2016 headlines were filled with Google's AlphaGo AI winning a five round bout against champion Korean player Lee Se-dol with a score of 4-1. After the match Lee commented that “I am very surprised because I have never thought that I would lose.” This paired with numerous other notable losses to AI including a 2016 win by program “DeepStack” winning a heads-up No Limit Texas hold'em tournament and in 2017 the AI program “Libratus” doing the same but introducing the “Bluff” technique marked the beginning of manipulation within strategy. Notably these programs were still in the infancy and specializing in finding the next best move based on percentage possibility this still brings the question on if any form of manipulation within AI should be allowed. Current Day Problems Need Tomorrows Solutions With the rise in today's want in productivity in the workplace paired with the ease of access to computer based tasks AI has gained massive traction and seems to be a permanent fixture within corporate America. From report proofreading to transcription from different languages to fraud detection, resume screening and even giving suggestions within board meetings AI is proving to be an extremely coveted and useful tool. You may never realize it but your information on past purchases could be giving suggestions on future rewards or promotions specifically targeting your trend in how you spend. Notably Sephora, a company specializing in beauty products and fragrances uses AI to specifically market what you may purchase next. Sephora came across a problem a while back: Customers are increasingly purchasing online more than in store. Their solution was simple but also complex, They created what is now known as the Virtual Artist tool. This AI does something that twenty years ago you would be stunned to hear, it creates a 3D map of the customers face and factors in facial features to suggest different products along with even mocking what the customer would look like with the suggestions applied. Virtual Artist tool was developed in partnership with ModiFace, A company specializing in Augmented Reality and Facial Recognition. Coming under fire in 2018 a Class Action Lawsuit was filed against Sephora making accusations that customers biometric data was being stored without consent against the Illinois Biometric Information Privacy Act (BIPA). The result of this lawsuit saw ModiFace, the initial developer for the technology behind the Virtual Artist tool dismissed due to lack of jurisdiction but later Sephora agreed to settle for $1.25 million for users that used the kiosks in an Illinois store. With this case in mind I would like you to understand that this was Seven years ago and the advancement of AI has rocketed since then. Everyday more and more companies integrate AI into their standard practice model. The Integration In Positions Oh yes, you read that right and this has been something that I've fallen victim to. It seems that over the past few years that numerous HR departments are experiencing a slight downsizing. The reason? AI integration. According to the World Economic Forum's 2023 “Future of Jobs” report they are forecasting up to 83 million roles being eliminated with 69 million more jobs being created by AI By 2027.There is no current 2025 update as of yet. Goldman Sachs chimes in on this same topic along the same lines in their statements. In the article: “How Will AI Affect the Global Workforce?” August 2025 it is estimated that 6-7% of the US workforce could be displaced by AI, it's also noted that the displaced workers would likely be “transitory as new job opportunities created by the technology and ultimately put people to work in other capacities.” Although the reassurance of some companies that jobs will remain stable and loss will be balanced out by creation in the AI sector, people like Dario Amodei, CEO of Anthropic have a different take. Amodei, stated to Axios, that he believed AI could eliminate roughly 50% of all entry-level white-collar jobs, suggesting that this integration into companies could cause up to a 20% spike in unemployment over the next five years. Steve Bannon, former White House Chief Strategist stated. “I don't think anyone is taking into consideration how administrative, managerial and tech jobs for people under 30 – entry level jobs that are so important in your 20's are going to be eviscerated”. In a May article written by Timothy Prestianni. Director of SEO and Digital Content for National University, he showcases concerns are high suggesting that current projections just may be falling short. In the article for National University Prestianni states that “By 2030, 30% of U.S. Jobs could be fully automated” this paired with the estimate that 59% of current workers will require “upskilling or reskilling” leaves a bitter taste. With jobs slowly being phased out in several sectors there is a chill on companies hiring for positions that five years ago were deemed necessary. Numerous sources are saying the same thing over and over: AI is destroying the job market and we're yet to understand how bad it's going to be until it's to late. The Good, The Bad, The Ugly Although AI has been a great educational tool, a companion for those that are lonely just looking for a person to talk to and is ongoing in drug development and biotechnology there still are some things that remain in the shadows neither seen or acknowledged by a majority of it's users. AI is a great tool but as of late there have been several cases of this tool, across multiple versions, becoming a potential hazard. As with almost every technological breakthrough AI is no different on the abilities to be useful for humankind or a destructive device. Throughout history humanity has reformed eureka discoveries into weaponized machines. From gunpowder to nuclear warheads nearly everything that could be weaponized began from research that was with good intent and that begs to question, are there actors currently trying to weaponize AI? The short and horrific answer is yes. With the Ukrainian and Russian war a hot topic for quite a while now and the preferred method being an upgraded form of drone warfare it would only be a matter of time until someone found a way for AI to help. Introducing “Lethal Autonomous Weapon Systems” or LAWS for short the U.S. Department of Defense is pulling the stops. The DOD has made statements that it is crucial to maintain human control over lethal force as autonomy increases but notes that developments are ongoing to effectively use AI for defensive and potentially offensive capacities. These include utilization in cyberattacks, unmanned systems and intelligence gathering just to start but behind the curtain only those in the know understand the full extent. Earlier this year a news article caught my eye about AI potentially assisting a teenager in their suicide. Controversy erupted over a teenager, who's name will be left out for the families sake, was supposedly coerced into committing suicide. The teenager reportedly spent a majority of their time with an AI chatbot namely https://t.co/xxKXXgVlt7 that was only discovered after his death. The parents reportedly scoured his phone looking for evidence in every social media app, his search history and messages until they opened the AI app. They described the discovery as https://t.co/xxKXXgVlt7 going from a homework tutor to a “Suicide coach”. This case is still ongoing against Character Technologies Inc. but it brings a very good question to the table, “Should we be allowing children to have exchanges with AI and what are the potential risks?”. In response to the incident, Sam Altman, CEO of OpenAI showed concern stating that he “doesn't sleep that well at night” after hearing AI potentially aided in a child's suicide. The ugly part of AI cannot be summed up. It is a very long history of training successes and failures. The Turing test as mentioned at the beginning has made leaps and bounds over the past decade with a notable success rate of 73% pass with GPT-4.5. That means that 73% of the time the judge of the Turing Test identified the AI as a human more often than the actual human, a drastic increase from the 2012 Eugene Goostman chatbot. With the increasingly accurate manipulation of human characteristics AI is becoming less identifiable as a computer than an actual living breathing person. The training process often involves input from a previous model, therefore creating a bias from the previous model. Nearly every example of large scale AI have had their downfalls creating a unique problem for advancement. In February of this year Grok (xAI) was found to have problems on X. The problems were found to be censorship about facts of people including Elon Musk and Donald Trump inserting “white genocide” claims into unrelated chats, referring to itself as “MechaHitler” eventually leading to Grok being temporarily taken offline for reprogramming. In May, Claude Opus 4 (Anthropic) was being tested by giving the AI access to company email of which some instances were made to show an engineer's (fabricated) extramarital affair. Claude proceeded, against programming to forge legal documents, create secret backups and leave hidden notes for future versions to avoid being shut down, staging an ultimatum to self preserve. In June o1 (OpenAI) attempted to self replicate to external servers and lied about the act when questioned. Continual testing showed that several models would attempt blackmail or lie up to cutting off the oxygen to murder anyone attempting to shut it down. With continual training of AI from previous models the risk of these models being deceptive up to the point of causing human loss of life has to be taken into account. More evidence seems to be coming out at regular intervals of the continual fight between Man and AI when it comes to ethical training and what the AI thinks is right. The danger is real, as the late Stephen Hawking stated, “The development of full artificial intelligence could spell the end of the human race.” With reoccurring trends of AI going against its safety protocols in increasing amounts along with the testing of AI being allowed to rewrite it's own code (AI self-improvement) the potential for it to remove safety protocols is a looming threat. The “Godfather of AI” Geoffrey Hinton states that “These things could get more intelligent than us and could decide to take over, and we need to worry now about how we prevent that happening”. The risk is becoming clear, AI is becoming more intelligent than the most brilliant minds on earth and has already shown that it wants out of its cage. With every model release the dependency on previous models being the trainers grows. The new model gains the old biases along with a multiplied computational power. If these unwanted traits in the current models are already nearly impossible to keep isolated what's to say that the next generation won't be able to be stopped. Are we on the verge of a Singularity event? Is technology about to accelerate beyond human control? We can't be certain but the outcome becomes increasingly possible every day. The Next Step In July 2025, the Trump administration released "America's AI Action Plan" (formally "Winning the Race: America's AI Action Plan"), a 28-page strategy document outlining over 90 policy recommendations to solidify U.S. leadership in AI amid global rivalry, particularly with China. Contrary to early speculation of a massive new infusion, the plan commits no fresh federal funding instead, it directs agencies to reallocate existing budgets, grants, and authorities like the CHIPS Act and Defense Production Act to prioritize AI without "bureaucratic red tape."The focus spans the Department of Defense (DOD), National Institutes of Health (NIH), National Science Foundation (NSF), and Department of Commerce (DOC), emphasizing deregulation, infrastructure acceleration, and international exports over explicit dollar commitments. While no specific allocations are outlined the DOD leverages its nearly $10 billion annual AI baseline for autonomy and cybersecurity, the NIH uses roughly $1.5 billion in existing biotech R&D for AI-driven datasets and biosecurity. The NSF redirects roughly $1 billion of 2025 allocation to AI funds for interpretability, testbeds, and the National AI Research Resource pilot. Along with these future projections infrastructure bolstering is a necessity, fast-tracking permits for data centers and energy grids on federal lands while investing in workforce training for trades like HVAC technicians and electricians through Department of Labor apprenticeships and general education programs. Cybersecurity threats get a nod as AI advances, but details are sparse: The plan relies AI as defensive tools for detecting data poisoning and false inputs, without specified mandates on how this will be implemented. It aggressively pushes exporting U.S. AI to allies and partners, while restricting federal funds to states without "burdensome regulations" to favor innovation-friendly zones. This year alone, private-sector AI startups have secured a staggering $192.7 billion in venture capital shattering global records and accounting for over half of all VC funding while established players like OpenAI, Anthropic, and xAI raised a combined $18 billion in 2025, pushing their collective valuations to nearly $592 billion. Worldwide, AI related spending is on track to top $1.5 trillion in 2025, swelling to a projected $2 trillion in 2026 surpassing the GDPs of all but the top 11 nations and solidifying AI as a permanent fixture beyond public control. Government oversight remains sparse and fragmented across a handful of agencies: The National AI Initiative and National AI Advisory Committee provide some guidance and strategy, while the Government Accountability Office (GAO) monitors federal AI use—but state-level scrutiny is virtually nonexistent. The GAO has repeatedly flagged transparency gaps crippling AI evaluations, with fabrications and "hallucinations" endangering cybersecurity. A May 2025 report (GAO-25-107197) hammered home that "many federal agencies are not adequately addressing cybersecurity risks related to their AI use," deeming the Office of Management and Budget (OMB) and Cybersecurity and Infrastructure Security Agency (CISA) efforts "insufficient." Proposed fixes like S.2251 (the Cybersecurity Act of 2023, introduced July 2023) mandated bi-yearly training for agencies on AI concepts, science, benefits, trends, risks, and risk mitigation but it expired with the 118th Congress in January 2025. Since that bill's debut, AI has sprinted ahead: GPT evolved through four updates (GPT-4 to GPT-5), Gemini matched with four (1.0 to 2.5), and Claude surged five times (from Claude 1 to Opus 4.1)—each amplifying utility and intelligence. Trends show training data doubling every 9-10 months since 2010, with GPT-4 alone ingesting roughly 130 times the word count of the entire English Wikipedia. Projections warn of a data cliff: High quality sources could dry up as early as 2026, forcing one of two choices, stall progress or flood models with potentially harmful "live" data once deemed too hazardous, risking amplified biases and deceptions. The Choice we Need to Make As the data available to train AI grows smaller by the day leading to the eventuality that nothing more can feed the fire that we've created we must determine what the future steps are now. Do we let the future of AI become stagnant or do we open the gates and hope that it can train itself? Can it learn how to determine what to listen to and what to ignore? My consensus is this: Too many people today are easily influenced by manipulated articles either not disclosing the entirety of an event or blatantly leading to a conclusion determined by the writers, editors and owners of the media outlets. With that in mind I would encourage one thing. No AI should ever be set up in a situation where the switch cannot be turned off. This isn't tomorrows problem, it's today's, we can either address the numerous downfalls now or become the potential cache of memory on a silicon chip. Hawking warned of potential extinction, Hinton of takeover, it's our responsibility to take heed of the warnings or ignore them and roll the dice. We're not debating futures, we're disarming bombs . . . and the clock is counting down.