The Hated One
patreon
Episode 261 - Sam Altman needs to be stopped
🕑 Added 2025-02-12 17:56:57 +0000 UTCThis is a work in progress for my upcoming YouTube video. If you don't want to wait for the video to come out, you can listen to this episode days before it's on my channel.
Intro
The duality of a tech bro – a benign Utopian with a dream to change the world. And a cut-throat CEO of a big tech startup looking to monopolize a new market all for himself. That is the story of any Silicon Valley idol in the past 40 years. And that is now the story of Sam Altman. The CEO of OpenAI, a former non-profit turned for-profit, that started with a promise to democratize AI, and ended with a total opposite of that. While in public Sam Altman is hyping up AI as the deliverer of abundance, behind the scenes he is consolidating power and concentrating the AI market. Instead of democratizing AI, Sam Altman is locking it down. Instead of open access, he is creating a black box.
To get to this point, Sam Altman used manipulation, lies and dirty politics, he surrounded himself with loyalists and discredited his critics. Now he has the ear of the most powerful politician in the world to effectively make OpenAI models exclusive default everywhere.
I want to build a case that will show you that there is no line Sam Altman wouldn’t cross to get what he wants. Some of these lines that he had already crossed are so disgusting it’s going to be very hard talking about them in a platform that immediately nukes any content barely mentioning these “forbidden topics”. That’s why I need your help – join my Patreon and support my work so that I can do more videos like this. I am very strict about sponsorship deals because I want to avoid conflicts of interests with my privacy content and that is very difficult. On top of that YouTube is just refusing to push my content to more people even if they are subscribed. So your help is an absolute necessity to keep me above water.
Stargate
Let me start from where Sam Altman is right now and how much power he has.
Sam Altman is a CEO of OpenAI, which still pretends to be a non-profit, but really is a for-profit and is basically a subsidiary of Microsoft. Microsoft’s entire operation now revolves around giving OpenAI all the infrastructure, money and resources to expand their AI into as many products as possible. OpenAI’s ChatGPT is practically the default AI in a rapidly growing number of companies, which now also include Apple and will soon include the US government too. That’s because OpenAI recently became part of a joint venture with the United States government, Oracle and SoftBank.
This joint venture will basically build giant water-guzzling data centers in US desert regions all on behalf of OpenAI and their partners. The venture had already committed $100bn and will rise the budget to $500bn over four years. [0]
The official statement is that this is “declaration of confidence in America”. [0] But the real reason this is happening, is that OpenAI haven’t found a way to make money off of their flagship models. [1, 2] The company is losing money on their premium subscriptions. [3, 4] And it doesn’t look like things are going to get better anytime soon because AI progress is being stifled. Not just OpenAI but the whole industry is running out of quality data and compute to train better models on. Chips, semi-conductors and other resources are more and more costly while giving back ever-diminishing marginal returns on improvement. [5] This problem has been apparent for quite a while. It prompted Sam Altman to call for the world to raise $7 trillion to build nuclear plants and data centers for him to train his AI on, but nobody took him seriously on that. [6]
Today, some big investors are questioning what trillion-dollar problem AI is trying to solve [5] when all it really delivers are models that confidently tell you to eat at least one rock a day [7] and beg you to die. [8]
With the threat of AI hype dying out, Sam Altman was probably going to run out of money very soon. The $0.5tn joint venture with Donald Trump is a divine salvation for him. But Altman will burn through that money very fast, because no matter how much money they throw at it, there is no hint in sight that would suggest AI development will not hit a ceiling. [9] The Internet has a lot of data, but quality data AI can be trained on is scarce. And synthetic, AI-generated data has proven to be poisonous to AI training to the point models trained on AI output deteriorate and even collapse. [10]
For Altman, this venture is a life support. For the US government – they’ve picked the worst AI company they could’ve possible partnered with.
Anti-open source
AI is still in its infancy and collaborative effort will be required to achieve meaningful breakthroughs. For such a young technology to progress, it necessitates open source and open access. Without it, independent researchers are not going to dig in and make improvements, when OpenAI could punish them for violating their license and terms of service.
The problem is OpenAI is the least collaborative tech company in the world. Everything they do is a black box. It’s not just that they refuse to open source their models. They refuse to share anything practical about how they build their models. So we know nothing about its methodology, data sets or model weights. The only research done on ChatGPT will be the research paid for by OpenAI. [11]
The reason Sam Altman refuses to open source his AI is because he thinks he can already make ChatGPT a commercial success and he wants to do it like Apple. He wants to create a walled garden of expensive AI tools everybody will have to pay for because they will be integrated into everything. It’s also similar to what Windows did to personal computing where it’s near impossible to go to a store and find a laptop that isn’t Windows, Apple being the only exception. Even though you have to pay hundreds of dollars for a Windows license, when you could get Linux for free.
Both Apple and Microsoft became dominant through a range of exclusive deals and the right lobbying campaigns and they completely crashed the system. But all that came at the cost of open access and it created market duopolies. That’s what Sam Altman wants and he is willing to become the enemy of open source AI in the process.
Open source fear mongering
If you listen to OpenAI, you will hear a lot of talks about safety, existential threats and risks. AI is too dangerous to fall into the “wrong hands” so by that logic open source is the arch nemesis. But OpenAI used to have a promise of open source AI. That’s what the word “Open” in OpenAI was meant for. The meaning of that was a lot deeper than just a software license – OpenAI was originally started out of fears that Google would have a monopoly on AI. [12] OpenAI was supposed to break that monopoly and prevent concentration of power within the AI industry.
But instead of becoming that, OpenAI began to copy Google’s strategy to concentrate AI. Instead of open-sourcing it, they made it proprietary. Instead of staying non-profit, OpenAI set up a for-profit company to attract bigger investors. In 2019, just four years since OpenAI was founded, Microsoft invested $1bn into OpenAI and acquired a 49% stake in the company, with a right to commercialize all past and future OpenAI technology.
So remember this lesson – if a tech bro sounds too good to be true, he is. Looking at it from today, Sam Altman never actually believed in open source tech. The fake promise of openness was a ploy to get billionaires like Elon Musk to donate money and get publicity. Now, OpenAI is counting its market cap in hundreds of billions.
Today, Sam Altman keeps fear mongering about why AI should still not be open source. Altman and his friends even went to US government on all levels telling them how very very dangerous AI is. [13, 14]
Everyone else’s AI, not OpenAI, of course. Their idea was to make it so that anyone who would want to develop AI that’s better than OpenAI’s GPT-4 would have to get a permission from the government to release it. A proposal which would effectively regulate open source AI out of existence. [14 – 16]
Luckily, no such law has passed, yet. Powerful Open Source AI is being developed by unlikely players such as Meta or China’s DeepSeek and many more and I am still waiting for the world to end. Maybe the world will really end when Google’s Gemini convinces enough people to add glue to thicken their pizza dough. [17]
DeepSeek fearmongering
Speaking of DeepSeek, I encourage you to pay close attention to what’s going to be happening in the coming months. [18]
Sam Altman didn’t hesitate to charge DeepSeek with the unforgivable crime of stealing from OpenAI, after OpenAI had been stealing from every creator and author on the planet without attribution or permission. [19]
Microsoft and OpenAI are saying they have “substantial evidence” that DeepSeek distilled knowledge from OpenAI models, which is a gross violation of their Terms of Service. [20]
As of this writing, they still haven’t provided any such “substantial evidence”. [21]
Look at the richness of irony in Sam Atlman’s tweet about this: “It is (relatively) easy to copy something that you know works. It is extremely hard to do something new, risky, and difficult when you don’t know if it will work. It’s also extremely hard to rally a big talented research team to charge a new hill in the fog together. This is the key to driving progress forward.” [21]
The rich part here is that OpenAI never charged the progress. Their work was built on top of AI research from Google, which was built on publicly funded and academic research. They are standing on the shoulders of the giants, but they are no giants. [19]
What thickens the irony even more, is that OpenAI accuses DeepSeek of distillation, something that’s a standard industry practice. Distilling is a method of using a larger model to train a new model that is lighter but has similar reasoning capabilities. [22] DeepSeek says they distilled larger models, but they say they used their own models, not OpenAI’s models.
Every AI team in the world is doing distillation. It’s what makes AI better, more customizable and accessible. The only reason this is even a news, is because OpenAI license doesn’t allow distillation to make competitive models. None of this would be the news if OpenAI was open source like it was meant it to be. But OpenAI is now on a brigade to “protect their IP”. [21]
What IP? Everything OpenAI has built only exists because of other people’s copyrighted works. Your models mean nothing without authors, artists and creators whose works you took without permission or attribution to build your own commercial products on top of. [23, 24]
What is your IP? Your argument works against you. If there is any intellectual property in ChatGPT, then ChatGPT should belong to every author of the content it was trained on. Just open source your AI and stop bitching about it. [25]
Even Zuckerberg open sourced his AI and that guy is literally AI. [26]
Manipulative liar
You probably know that in 2023, Sam Altman was actually fired from OpenAI by the board. Only for a brief period of 5 days and was re-instated. [27, 28]
What you probably didn’t know, was that the reason was that he went rogue and was on a path to transform OpenAI into an AI dictatorship.
The original board members were the people that feared powerful AI, but they didn’t just fear it for the hypothetical scenarios of a rogue AI takeover. Those AI fears were there, [29] but they were also concerned with such a powerful AI falling into the wrong hands. For the longest time, the effective altruists on the board thought the public were the wrong hands. But Altman made them realize that they could also belong to one of their own.
When Microsoft bought the 49% stake at OpenAI, Sam Altman lied to the board. That connection didn’t just come with the money. [30] It also came with personal protection for Altman from Microsoft CEO Satya Nadella, with whom Altman had a close relationship. The board saw how Altman quickly transformed OpenAI from a safety-first non-profit to a move-fast-and-break-things for-profit company. [30] Altman pushed for a fast-tracked release of ChatGPT when Anthropic elected to delay the release of their model specifically to slow down the AI rush. [30, 31]
One of OpenAI’s board members, Helen Toner, wrote a paper criticizing Altman’s decision to “stoke the flames of AI hype” with the “blockbuster release” of ChatGPT, spurring “frantic corner-cutting”. [31]
When Altman heard about Toner’s critique, he went rogue. Instead of admitting to his mistakes, he began scheming with other board members, trying to get Toner removed from the board. [30] Even when many disagreed, Altman would lie about their opposition to make it seem like there was a consensus to get Toner replaced. There wasn’t. [30]
Altman was becoming more and more dictatorial and the board began to realize they actually couldn’t hold their C.E.O. accountable. [30] Which is the only job the board has to do. So, four of the board members, Toner included, decided to vote Altman out of his position. But they were too afraid to do it openly, fearing that if Altman had found out, he would have done everything he could to undermine the board. They were determined to catch him off-guard. [30]
But when the vote came in and Altman was ousted, Microsoft got involved. They threatened the board that if they don’t reinstate Altman as C.E.O., they would move him and all of his loyal employees over to Microsoft. OpenAI would be left with nothing. [30] Rather than facing the prospects of a total failure, the board obliged with Microsoft’s request and all board members except for D’Angelo resigned. [30] As a bonus, Microsoft gained a nonvoting board seat at OpenAI. Sam Altman won everything. He got the money, the team, all the IP… and there would be no one left standing in his way. OpenAI is all just Microsoft and Sam Altman’s loyalists. [30]
Cheap labor
One of Altman’s best kept secret is hidden behind the lie that ChatGPT is a testament of computer thinking and artificial intelligence. What isn’t told often is that ChatGPT was built on top of cheap labor in poor countries. [33]
To make ChatGPT less toxic, Altman needed to teach the model to reject prompts generating malicious or illegal content. But computers can’t do that well because they don’t recognize the nuance of human language. [32] So to achieve an investor-friendly AI model, OpenAI began to feed it labeled examples of violence, abuse or hate speech and hoped the AI would learn to detect those and filter them out. [33]
Building such a detector is most effectively done by hand. Which means thousands of people have to go through millions of hours of the most horrible content you can imagine. All the naughty words such as harm, abuse or murder I can’t even mention because YouTube thinks you are not mentally capable to handle it. I am once again asking you to join my Patreon and support my work for as long as you can.
OpenAI hired an outsourcing company called Sama, that employed workers in Kenya for as little as $1.32 per hour after tax. [33] Even for Kenya that wasn’t a lot, because it was less than the lowest-paid receptionist in Nairobi would earn. OpenAI contracted Sama for 8 months, but the collaboration collapsed within weeks since it started. [33] The work OpenAI requested from Sama was so traumatic to the workers, they described the OpenAI assignment as torture. Employees described being mentally scarred by the work so much they even found wellness counseling insufficient. But, somehow, the situation gets even darker from here. [33]
When the contract with Sama started, OpenAI began sending additional instructions for “some illegal categories”. [33] Among them was a request for Sama to collect the most illegal and immoral images you can think of, and send them to OpenAI. And yes, that includes CSAM. [33] Sama provided a sample of 1,400 such images for which they charged OpenAI $787.50. [33]
Now, using cheap labor to improve AI models is not unique to OpenAI. Facebook is also using sweatshop labor to label illegal content. But OpenAI sending requests for illegal content was a bridge too far even for Sama and they decided to cancel all of its work with OpenAI. For some workers, that would also mean losing the $70 explicit content bonus per month. [33] The OpenAI contracts were so disastrous for Sama the company made it their official policy to stop all work for OpenAI, and ceased all work on natural language processing and content moderation. [33]
The most scandalous aspect of this story IMO, is that these OpenAI contracts only cost $200,000. OpenAI could have easily afforded to pay far better wages and safer working conditions for their workers and it would barely make a dint in their budget. [33]
Data? What data?
This story also hints at what kind of data OpenAI trains their models on. Because it clearly involves very illegal corners of the internet. Which begs the question – how did you get this data? Where did you get it from? Did you deliberately go these illegal websites? Or did you get them from people’s private messages on social media? If you are looking for answers on these questions, Sam Altman categorically refuses to give any meaningful answer. He defaults to corporate vaguespeak which says less than saying nothing at all. [11, 34, 35]
OpenAI admits they train on publicly available data, which for instance includes public YouTube videos. [36]
I noticed that at least one of my videos ended up in some AI training data. [37] No, I didn’t consent. OpenAI also gets data, supposedly not so publicly available, by reaching partnership deals. [38]
What data? Does that include our chats, emails, unencrypted calls and texts? How did ChatGPT learn how to text like a human?
This is the biggest reason why all AI should be open source. Sam Altman took all of our information, public and personal, used it to train an AI that wouldn’t exist without our data, and then said that AI belongs to him and him alone.
God-Altman
One thing I want to highlight is how much of a political weasel Sam Altman is. This is important because politics is power and Sam Altman is an opportunist. Here is young Sam in 2016, criticizing Trump as an unacceptable threat to America, saying he wants to stop him and vote against him. That was also in 2021. [39]
And here he is now praising Trump as incredible for the country. That is, of course, bought and paid for admiration, because who wouldn’t love Trump if he backed your $500bn venture. [40]
And here’s Altman embracing MAGA loud and clear. Oh, sorry, it’s MSGA. [41] Because he is talking about search. Not America.
I am not saying that people can’t change perspective, but going from an “unacceptable threat to America” to “incredible for the country” is a pretty big leap. Altman is doing this because Trump will help him cement OpenAI as an incumbent in America. Thus finally succeeding in dismantling Google monopoly, only to replace with his monopoly. Power is only offered to those who ready to lower themselves to pick it up. And there is no low Sam Altman wouldn’t go to to pick it up.
Like any good tech bro, Altman is also building his own messianic cult. He is presenting himself as the visionary that offers abundance and technology to solve world problems. Here is Altman calling AI abundance. [42] This is a lie, fake news, disinformation.
AI is the most resource-constrained tech ever. An AI assisted search is 5-times costlier than a conventional search. AI is dramatically driving the demand for water and energy, completely destroying all the net zero targets big tech promised in their green washing campaigns. AI also doesn’t exist without rare minerals which are extremely scarce and only available at few geopolitically tumultuous locations. ChatGPT is not abundance. It’s the opposite of abundance. [43. 44]
But Sam Altman needs to hype up AI because without the hype, it doesn’t have the value that would justify its costs. Here is Altman selling abundance as a solution to reduce conflict. Maybe we should have done that all along! Hey, Israel! I know you are in an ethnic war at the moment, but have you tried abundance? Hey, Putin, you said all you wanted was a land-bridge to Crimea but did you know that abundance? Hey, drug cartels… abundance! Abundance, anyone? Abundance! Altman wants to plant this seed in your mind that AI = abundance = world problems and conflicts solved forever. That’s a sci-fi scenario that has nothing to do with the value brought forward by generative AI like ChatGPT hallucinating fiction about literally anything if you just keep asking.
Conclusion
I wanted to bring you this record specifically about Sam Altman, because nothing is more harmful than giving manipulative personalities powerful businesses to run and a global political prestige. His goal is to hoard AI technology for himself and he’s making deals with businesses and politicians to restrict others and elevate himself. He somehow convinced Microsoft and Apple, the most powerful corporations in the world, to bow before him. ChatGPT is now exclusive AI on the entire platforms of Microsoft and Apple with no relevant competition. Google doesn’t even compete with OpenAI anymore, because Gemini is now almost exclusively on Google products and Android. Sam Altman is a rent seeker and he will not stop until OpenAI is an absolute monopoly.
OpenAI doesn’t yet have that monopoly. Its biggest enemy is not Google but open source AI. Today, thanks to Meta’s Llama models, DeepSeek, Mistral and many other projects, there is plenty of open source AI out there that’s just as capable. Sam Altman wants to stop it and he will fear monger about them. He is already doing it to DeepSeek. But don’t let him make that decision for you. Use open source AI. If you can run it locally on your device, do it. If you can’t, running it in the cloud is still more beneficial in the long run, even though it’s not private.
My ultimate position is that technology should belong to everybody. I am not against commercialization, but I am against IP hoarding. Technology is built collaboratively and should be given back to the people.
I also want you to know that I am principled and I want to make content to the best of my ability and conscience. That’s why I ask you to join my Patreon and support my work so that I am able to do it as long as possible. I am only open to sponsors that wouldn’t introduce a conflict of interest with my privacy content. Which does hurt me financially, but I think ethical content matters more. Please support my work as much as you can and you’ll be greatly rewarded with hundreds of podcast episodes I’ve been making on patreon.com/thehatedone. Thank you.
SOURCES
[0] https://www.ft.com/content/48eb53a1-67ca-4509-8c62-401f0cf8b099
[4] https://twitter.com/sama/status/1876104315296968813
[7] https://www.boredpanda.com/google-ai-overviews/
[8] https://gemini.google.com/share/6d141b742a13
[10] https://www.scientificamerican.com/article/ai-generated-data-can-poison-future-ai-models/
[11] https://cdn.openai.com/papers/gpt-4.pdf
[13] https://www.politico.com/news/2023/10/13/open-philanthropy-funding-ai-policy-00121362
[15] https://news.yahoo.com/google-brain-cofounder-says-big-113049941.html
[17] https://www.bbc.com/news/articles/cd11gzejgz4o
[18] https://www.cnbc.com/2025/01/27/chinas-deepseek-ai-tops-chatgpt-app-store-what-you-should-know.html
[19] https://www.404media.co/openai-furious-deepseek-might-have-stolen-all-the-data-openai-stole-from-us/
[21] https://www.ft.com/content/a0dfedd1-5255-4fa9-8ccc-1fe01de87ea6
[22] https://www.ibm.com/think/topics/knowledge-distillation
[23] https://www.wsj.com/tech/ai/scarlett-johansson-openai-sam-altman-voice-fight-7f81a1aa
[25] https://www.nytimes.com/2024/05/29/technology/what-to-know-open-closed-software.html
[26] https://llama.meta.com/llama3/license/
[27] https://www.ft.com/content/46efa770-4b47-49bb-b0f8-824f1c4f38a3
[29] https://www.straitstimes.com/business/effective-altruism-s-role-in-the-openai-drama-explained
[30] https://www.newyorker.com/magazine/2023/12/11/the-inside-story-of-microsofts-partnership-with-openai
[31] https://cset.georgetown.edu/publication/decoding-intentions/
[33] https://time.com/6247678/openai-chatgpt-kenya-workers/
[35] https://openai.com/policies/row-privacy-policy/
[36] https://www.wsj.com/tech/ai/ai-training-data-synthetic-openai-anthropic-9230f8d8
[37] https://www.proofnews.org/youtube-ai-search/
[38] https://openai.com/index/data-partnerships/
[39] https://twitter.com/VinhLeRealty/status/1882268241621586299
[40] https://twitter.com/sama/status/1882234406662000833
[41] https://twitter.com/sama/status/1887431071190396939
[42] https://twitter.com/sama/status/1452336044691632129
[44] https://www.nature.com/articles/d41586-024-00478-x
Comments
sonnysighedup
Thanks for covering this and for all of your other work. I was wondering if you had looked into the case of Suchir Balaji, a former OpenAI researcher and potential whistleblower, who was found dead in his apartment last November in an apparent suicide. Some links of interest: His personal home page: https://suchir.net/ A NYTimes piece about him from October: https://www.nytimes.com/2024/10/23/technology/openai-copyright-law.html An interview with his mother by Tucker Carlson about the suspicious circumstances around his death: https://rumble.com/v69sarg-mother-of-likely-murdered-openai-whistleblower-reveals-all-calls-for-invest.html?e9s=src_v1_ucp
