I think it’s means he’s one of the creators of it, which as the CEO of the company is actually pretty fair. It’s the same sense in which Bill Gates created Windows XP, for example. Probably wrote zero lines of code for it but it’s his company and he was probably a huge part of the vision. CEOs often are co-creators of things, especially in smaller, tech-focused startups.
“ChatGPT creator, Sam Altman” ? or “OpenAI CEO, Sam Altman” ? Two very different things, he didn’t create ChatGPT, please give the credits to the right people.
Yes he is absolutely is one of the creators of ChatGPT. It’s so silly when people act like only people with hands-on-keyboard can be considered creators.
@@therainman7777 no he isn’t, here what some employees think about him: German broadcaster Bayerischer Rundfunk (BR24) reports that OpenAI is losing key employees to Google. Some of these employees have already resigned and signed contracts with Google. Others will do so in the coming days. The disgruntled employees are reportedly unhappy with the development of ChatGPT and the rapid growth from about 100 to 600 employees since December 2022. Sam Altman, CEO of OpenAI, is criticized for having only a "superficial understanding" without much involvement in day-to-day
"We're not training GPT-5 right now" He didn't say they weren't working on it. In fact, the actual training process only takes a few days. What takes time is prepearing the training data
I'm not sure where you got that number from lol. Deepfake models take days to train, GPT4 took over a month. GPT5 most likely similar if not more, if there is an increase in parameter count. The pretraining and data gathering does take a long time, but I have a feeling that chatgpt and a lot of data that's currently being collected are going to contribute to training gpt5.
@@rasuru_dev 6 to 8 months was for Reinforcement Learning with Human Feedback. That was done after the training of the base model. They never revealed the training time for GPT 4 but GPT3 was trained in 34 days. Assuming they trained with a similar number of epochs for GPT4 then it would be perhaps a couple weeks longer than GPT3. With the immense dataset and parameter size, it would have been extremely expensive to keep training it for longer than that. Plus, you get diminishing returns the longer you train. RLHF is ultimately where the magic happens.
I agree with Sam in that you still need to follow the proper business guidelines. However, in terms of what the target product is, I believe it’s important to take a giant leap toward what you think is reasonably possible within the new framework. For me, I’ve seen this within video games, when there is a big platform a shift, many of the losers are still chasing the old goals, and the winners find the new capabilities that are not obvious until they are in hindsight.
@@test-zg4hv Well it is different but, video games are a *multi-billion* dollar industry, in fact has been the *most profitable* entertainment industry for years now surpassing movies, music, sports, and any other entertainment form you can think of, in 2022 the worldwide gaming market was estimated at almost *347 billion* . Video game industry: 347 billion USD Film industry: 43 billion USD Music industry: 23 billion USD Sports industry: 73 billion USD
Of course they are different, I wasn’t implying that they are the same. Just speaking of personal experience about people and their use of things that are sizably different than in the past. The early days of game development does, however, have some overlap. Carmack is a great example. He loves new challenges, so he left Oculus and is now working on AI.
From video, working on AI, some guidelines for all: Never loose site of vision and focus and vision on long term. But a very tight feedback loop on what is working and what in not working and doing more stuff of what is working and less of what is not working and very very careful user observation can go super far Why CGPT successed? Beig on ground for long time and sweating every details, most people aren't willing to do that
Six days ago? Didn't I watch this on a different channel almost two weeks ago? Dang, it's really hard to keep track. 90% of all these Sam interviews are so similar. It was only when Lex got up that I instantly remembered I'd seen this already. Is this copy/pasta content on Forbes?
@@ghost_mall you mean a machine that expedites all information by the millisecond running thousands of prediction and doing in 2 seconds what it would take you a lifetime? And you think it can't take existing information and do the very same. It's not the machine it's the person building it.
@@ghost_mall not training though is it, relevant information however. It's all one and zero. That makes no difference. Every question asked and answered on ChatGPT 1 is false.
What's so slow about these neural networks, what are you doing there? Everything served on a tray and you mess with it like a girl in front of the mirror getting ready to go to a party. You don't get it or what? I've seen the schematics good direction, Libra, you have everything, what is it about? Because the sooner the sooner it will reach you to apologize and fix what you broke and bear the consequences. respect+position+trust
As someone that's used ChatGPT literally too many times to humanly count, with dozens of active, continuing conversation agents, for multiple different purposes, you're absolutely stoned off your gourd if you think GPT-4 can write "an excellent book." No amount of prompt engineering can get it to produce genuinely good work, what it produces is, at best, above average. Sure, that still looks AMAZING to the average norman Andy with little to no reading comprehension beyond basic, American public school-level literacy skills and a probably deficient imagination to boot, but GPT-4 sure af ain't producing anything appealing to anybody with a cognitive capacity beyond that of the average centrist, conservative neoliberal American. Maybe all it needs is a sufficiently massive token count that it's capable of producing actual, coherent chapters instead of just 3-4 paragraphs of centrist virtue signaling at a time. Conversely, tasks involving the Code Interpreter and without the need for a creative spark/soul can be done with BRILLIANT efficiency, assuming you know how to prompt even half-decently, and you can get fantastic results with some of the cutting edge prompting research that's available. Alas, GPT's context window and lack of meaningful memory prevents it from producing anything narrative of any meaningful length and novelty.
I am probably in the minority, but there is something genuinely un-likable about Sam. Can’t put my finger on it, but he seems very far from genuine in his interactions.
GPS overwrites Your Mental map you use to have. And ability to find your own Routes. Chat GPT overwrites your research ability to aproach topics in different angles then anyone else. Every rabbit hole has many choices of paths inside it to test other hypothesis or disprove past "facts"...ChatGPT is here to narrow that rabbit hole path to one tunnel...to the answer it gives you...a propagandas dream tool...cant have people REASEARCHING TOPICS THEMSLEVES ...can we
Tell me how you research things yourself currently with existing tools. Is it the same way you've done it in the past 10 years? Hint: if you've literally used any search engine your focal window is already being routed by ai. Sorry bout it, maybe blame certain far right leaning political oligarchs.
Hey Sam you need to laugh and smile. Plus don’t be so elusive when questions are asked. Well he will be not as important anyway in late 2024 once agi is achieved. As is in the next few months ai will be much smarter than almost every human in the world
ChatGPT has its pros. However… It certainly can affect cognitive decline because you’re not actively using your brain to think. It can make things easy. And activates other unhealthy habits like giving away our power to use critical thinking. Technology is fascinating, helpful, and has cons if not use properly. We can only hope for the best for the good of all. ❤
Or it frees up mental space by doing a lot of the mundane so I can focus on the interesting. It’s all how you use it. Just like RU-vid. One can watch fail army all day or one can learn a new skill. Your choice.
Here is my hypothesis which was recently confirmed by our team: In collaboration with the State University of New York (SUNY), Massachusetts General Hospital, Tokyo Institut of Technology, and University of Munich, I have been conducting research relating to three modes of cognition ― instinctive, intellectual, and intuitive. MRI Study of the importance of Instinct, Intellect and Intuition in Human Intelligence Introduction Understanding intelligence is one of the great scientific challenges of our time. In recent years, many groups have made progress measuring brain activity using numerous non-invasive methods and a series of notable studies have been made by several research teams. While the relationship between the work of neural circuits of the brain and intelligence seems obvious, the hypothesis of ‘neural efficiency’, which was long used as an assumption by most neuroscientists and psychologists, was confirmed experimentally (Smith et al., 2015). Studies of human intelligence provide strong evidence for the neural efficiency hypothesis, which suggests more efficient brain functioning (i.e., less or more focused activation) in more intelligent individuals. Additionally, the role of intuition versus deliberation in human cooperation was studied (Bear & Rang, 2016). The latest study reports the first evidence that electrically stimulating the dorsolateral prefrontal cortex can enhance our ability to ‘think outside the box’. (Luft et al., 2017). Another few groups of researchers demonstrated that functional connectivity profiles act as a ‘fingerprint’ that can accurately identify subjects from a large group (Finn et al., 2015). The results from this experiment also indicated that an individual’s connectivity profile is intrinsic, and can predict levels of fluid intelligence. The variation of brain functioning in different people should be examined. The proposed study allow to avoid overly complex methods and might help to identify the principles of intelligence and fundamentally advance our understanding of intelligence as a whole ― be it human, animal, individual, or collective. Method MRI can show the activities of the specific brain centers in each of three groups of participants and their brains’ wiring patterns. 1. Illiterate Individual who are not exposed to formal methods of schooling and thought. 2. Educated individuals, who are trained in thinking logically and understanding complex ideas and concepts. 3. Leaders of science, technology, business, economy, and culture with proven ability to use their intuitive insight. Instinctive mind: The first level of cognitive mode is instinct: the way in which people naturally react or behave, without having to think, which is generally understood as an unconscious mental process. Instinct is directly related to the ancient, animalistic group memory which is responsible for survival. It is below the level of awareness and activates automatically. The brains of such pure instinctive individuals should have some specific centers responsible for the operation of instinctive mind and detectable wiring patterns between them. Intellectual mind: The second, higher level of cognitive mode is intellect, a conscious mental process with a self-learning neural network. Intellect is responsible for the accumulation, organization and preservation of knowledge from the past. It is based on the accomplishments of the past, the repetition of the past and closely related to a collective mankind memory. Intellectual thinking consists of deduction or inference, which are the rational abilities of the mind. But intellect works only in a linear direction - from the realm of the known to the known. "The intellect selects in a given situation whatever resembles something it already knows, or is familiar with; it seeks this out to apply its principle that 'like produces like' " (Bergson, 1907). Intellect is incapable to produce ingenious ideas. Nevertheless, intellect has two important roles: First, it is responsible for translating intuitive discoveries into scientific or cultural forms understandable to the general public. Second, intellect has a fundamental social function - it unites humans with other humans. Intellect employs some of the following conscious mental processes: self-reflective thinking, planning, predicting, deliberation, attention to or monitoring of their actions, conceptualisation of action, control, effort, and acting for a reason. Compared to the instinctive mind, intellectual mind has higher level of awareness and self-awareness and more features of decision-making. The brain of such intellectuals should have detectable specific centers responsible for the operation of intellectual mind and wiring patterns between them. Intuitive mind: The third and the highest level of cognitive mode is intuition or, in other words, intuitive mind. It uses the most advanced neural network and requires the highest level of awareness and self-awareness. Intuition is not analytical and is not to be confused with instinct; it is above both ― the animal instinct and rational thought. In comparison to instinctive and intellectual minds, intuitive mind is NOT based on the ancient/animalistic groups’ memory or collective intellectual memory, as it does most of its discoveries contrary to established collective patterns and regardless of the views of the highest authorities. The brain of such individuals should have detectable specific centers responsible for the operation of intuitive mind and wiring patterns between them. BUSINESS PROPOSAL onedrive.live.com/view.aspx?resid=870D7D1389B4E11E!7585&ithint=file%2cdocx&authkey=!ANWA7lhZDS5nc1A
