ChatGPT Is Changing Our Understanding of the Human Intellect
Vložit
- čas přidán 9. 05. 2024
- ChatGPT is changing what it means to feel uniquely human. Large language models have the ability to mimic human creativity and spout human knowledge at record speeds-and they’re getting more powerful by the day. But while this technology might seem scary, it doesn’t need to be.
Every CEO Dan Shipper puts it this way: “ChatGPT doesn’t replace me. It just changes what it is that I do.” In this essay, Dan makes a case for a new understanding of human intellect, drawing on examples from medicine, psychology, and philosophy.
If you found this essay interesting, please like, subscribe, comment, and share. If you want more from Every on cutting-edge technology, check out our weekly podcast series, How Do You Use ChatGPT?, where Dan interviews luminaries and experts like Reid Hoffman, Tyler Cowen, and David Perell.
To hear more from Dan Shipper:
Subscribe to Every: every.to/subscribe
Follow him on X: / danshipper - Věda a technologie
Even if AI doesn't get phenomenally better, its power lies in removing the repeat work we've done billions of times as a species, particularly data based work. We've made millions of forms, millions of documents, millions of stories and millions of programs over the last 4000+ years (Moreso in the last 70 or so) and what the latest AI's are is a compressed store of that knowledge and skill that it took to make those documents, forms, stories and programs. Instead of doing the same work over again we have a means to communicate with the data in a natural language and receive data back that's largely something that's been done before though with different names and variables. It's an incredible tool.
This is great and summarizes 😉 a lot of ideas I’ve been having about LLMs and how they’ve changed how I think about myself and how I think in general. I find the skill I value most now is knowing the right questions to ask, knowing how to frame questions, and understanding what the audience I make content for wants or needs to hear. It makes me think of ideas by their context, framing, and connections to other ideas more than the content itself. Bravo 🎉
Brilliant Dan - this was fantastic. Glad you created this ‘essay’ and I look forward to you fleshing out what it means to be uniquely human.
First off, I want to commend you on an exceptionally well-crafted video. The depth and nuance of your discussion around AI, creativity, and the human intellect is truly engaging and thought-provoking. Thank you for sharing your insights.
I'd like to offer a couple of perspectives that I believe can further enrich this inquiry:
While understanding AI's current strengths in summarization and information synthesis is important, we should recognize that this is but a glimpse of its potential. Soon, AI will likely evolve into a far more advanced thinking machine, capable of profound analysis and ideation beyond mere technical organization.
More crucially, I believe the essence of human creativity lies in our ability to create from a place of not-knowing, from an empty space of pure potential. True innovation springs from this source, rather than from the recombination of existing knowledge, which is inherently limited. AI's real gift to humanity may be in liberating us from the confines of mechanical thought patterns, freeing us to dive deep into the infinite wellspring of creativity.
In this light, relying on intellect as the key differentiator between human and machine is a losing proposition. To truly harness the potential of AI, we must approach it with a beginner's mind, letting go of preconceptions and allowing the radically new to emerge.
I believe that by embracing this perspective, we can steer the evolution of AI in a direction that uplifts and empowers our most profound human capacities.
Thanks again for the fantastic video and for sparking this vital conversation.
what we are = what we know is already placing ourselves in the space of the machine. My kids for instance are learning lots of stuff, but it's much more important how they learn, and why they learn than what they learn…
This feels documentary-level. So excited for this
I just started asking Meta AI about tokens and I learned more in a few minutes than I have in a month trying to figure out this stuff on my own. I've used it to write SQL code in seconds that would have taken me at least an hour. Maybe they're just "summaries", but it's crazy how fast these LLMs are improving so maybe you should revisit this topic in a year or maybe less lol
One thing i have noted in my interactions with ai is how they can act as a sounding board to help develop ideas. Like having a friend to brainstorm with
Well put and articulated. Thank you.
What defines us as humans is not our intellect, but our ability to reflect upon our actions and experience shame or guilt when we make a mistake, or pride and joy when we do something commendable. In other words, a conscience…
Animals operate on impulse. Less intelligent humans do as well. More intelligent people reflect.
Totally ridiculous and archaic ideas. Humans are animals and all humans operate on instincts as well. Falling asleep when you're too tired is a form of instinctive behavior. Sexual arousal is too.
Animals feel shame, joy, empathy and so on. It's obvious for anyone who has being humble and intelligent enough to see the natural world without sickening anthropocentric superiority.
What makes us humans is nothing more than a slightly higher IQ than other animals. Nothing else. Nothing special.
That's why so many of our language's benefits were so easily reproducible by LLMs.
Even parrots can talk like humans. Nothing special about human language.
If you think it's special or that you're for talking it, try to comunicate with an ant. Let's see if it's so easy. I mean, after all, you're "superior" to them, right? 🙄
Teach the AI how to manipulate Humans, you're going down.
Humans are animals and all humans operate on instincts as well. Falling asleep when you're too tired is a form of instinctive behavior. Arousal is too.
Animals feel shame, joy, empathy and so on. It's obvious for anyone who has being humble and intelligent enough to see the natural world without anthropocentric superiority.
What makes us humans is nothing more than a slightly higher IQ than other animals. Nothing else. Nothing special.
That's why so many of our language's benefits were so easily reproducible by LLMs.
Even parrots can talk like humans. Nothing special about human language.
If you think it's special or that you're for talking it, try to comunicate with an ant. Let's see if it's so easy.
I think what makes us humans is a combination of all aspects of us. Boring answer, but true.
I would agree. Kind of sucks right? That intelligence is actually decoupled from consciousness. It seems intelligence is the dot product of pattern recognition and association. Very interesting thought experiment. Will future systems inherently be sociopathic? What about alien life? I don’t like OpenAI, but they definitely caused a thought shift in so many ways.
Thanks - that was good for me. I'm an Ai consultant and sometimes it's hard to explain what makes us human and Ai Ai. You gave a beautiful suggestion and direction in which this discussion could go.
Really glad you liked it!
from time immortal, people have feared change, new ideas or new thoughts....change happens...evolve or die...
I was drawn to the title. ❤🎉. Stephen Wolfram wants LLM science
I feel like your advice works for now, but I wonder how future proof it is given the unending upgrades AI is experiencing.
Wait wait. Why is one of the most insightful , calm and discursive of AI podcasts down to 5.8k subs?
It's insanely good, like if Terry Gross had a baby with Charlie Rose and that baby lived with Tibetan monk for ten years but then they had a climactic fight on the Anapurna and the monk said before falling down the cliff 'I am your real Father. Terry and I were lovers' so the boy ran down the trail in tears and at the bottom he discovered a cake and coffee shop where he was handed an iPhone and then said Lets see what this does, then started a podcast...
this is without a doubt my favorite comment ever. thank you!
I was drawn to the title. ❤🎉
Donning my Summarizer Hat: this viewpoint is essentially the "stochastic parrot" argument, but with a positive spin. It's likely incorrect, or (more charitably) misleading. Contrarians notwithstanding, the ML research community mostly dismisses this perspective. The gist:
1. Perhaps intelligence essentially _is_ compression (see Hutter). Perhaps analogy _is_ reasoning (Hofstadter). Regardless, the point is that we cannot extricate summarization from creativity, construction, generalization, memory, etc. Every word you write, and every skilled artistic paint stroke, is the amalgamation of others' work. The sentence structures from children's books, the stylistic expression of Monet, the turns of phrase spoken by peers -- it's all rooted in "summary." Humanity sans "summary" might be a meaningless notion.
2. LLMs exclusively extrapolate, _never_ interpolate (LeCun et al., 2021). In other words, SotA chatbots don't glue together bits of training data. Their output is always "creative." This doesn't imply that LLMs _can't_ reconstruct training data, anymore than a human can't reproduce e.g. their national anthem's lyrics. But both systems get there through constructive generation.
3. These frontier models have _far_ too few parameters (on the order of GBs to a TB) to represent their breadth of knowledge and skills through regurgitation. They simply don't have enough bits, by many orders of magnitude.
4. LLMs have phenomenal zero-shot and few-shot learning capabilities. That is, they can consistently reason about concepts and data which they've never seen before.
5. Transformers and deep neural networks are, well...transforming (sorry) nearly every domain. We've effectively solved protein folding, with corresponding new drugs on the way. Machines have conquered Go, an "impossible" feat (unlike chess) due to the game's astronomical solution space. Robotics has experienced a revolution (see e.g. RT-2). Mathematical simulations with DNNs have seen at least a 10x acceleration, unlocking an explosion of scientific progress (e.g. fusion tokamak leaps). The AI-fueled explosion of progress is so vast outside of the narrow scope of chatbots that I couldn't possibly do it service.
18 months ago, I couldn't get Siri to acknowledge me half the time. Now, I can snap a photo of my refrigerator's contents, and GPT-4 will suggest a recipe, give directions, and warn that I'm nearly out of milk. If this is merely better information access with Google v2.0, just summarization, then maybe _so am I._
I appreciate your video, friend. It's well-written, insightful, and genuine. I wish that I shared your AI tranquility too. Cheers.
Really great writeup. An interesting point to consider, however -
If you ask chatgpt who Tom Cruise's mother is, it will tell you Mary Lee Pfeiffer, which is correct. However, if you then ask in a different context who Mary Lee Pfeiffer's son is, it will hallucinate an answer. I just tested that this still works even with GPT4.
My hypothesis is that this happens because on the internet, you will rarely, if ever, find Tom Cruise referenced as his mother's son. It will almost exclusively reference her as his mother.
That said, there is clear evidence SotA models are doing some sort of reasoning. For example, training on code improves image recognition capabilities.
4. They're not believed to be reasoning. They only identify a pattern on their training data that leads to an answer and they learn key points that identify the same pattern in other data so they can extrapolate the answer, with no need to reason about it.
Sometimes it seems like they're doing some impressive mental work, but then you talk more with the model and you just see nothing there. Just pattern matching. Feels sad tbh. But I still believe that AI can achieve AGI and sentience. Someday.
even calling it „A.I.“ is misleading, these programs dont have any „intelligence“ at all
still AS - artificial stupidity
and Altman is correct, the dumbest of all is GPT4
LLMs don't think, they are a model of language. AI is a bit of a misnomer, this is not intelligence, but rather an application of machine learning, which is simply a computational process of creating models based on the measurement between points of information in a large dataset.
The anxiety most people are feeling is not existential, it's economical. It's not the technology that's scary, it's how other humans will use it that's scary. This technology is being used to extract more value from us while compensating less. People are not afraid of AI, they're afraid of capitalism.
Within that fear, I think a lot of people are missing the actual revelation of LLMs. Defining language has always been a tautological problem, like trying to describe water when the only word you have is wet. LLMs are an empirical tool that we can use to clearly define how language functions and how the meaning of words are actually derived.
Language is one of the most mundane things, every human in history learns it instinctually and employs it on a daily basis, but this is the first time in history that we've been able to mechanically emulate language, which is absolutely profound.
Well put! I was hoping the video would be something along these lines. This video “looks good” but its content is more shallow than it seems. it’s helping people cope with the fact that we’re a little less special, not a deep discussion.
@@chasealcorn1047 Yea, he doesn't really even attempt to define intelligence, claiming that it's "fuzzy".
Reminds me of a teacher I had for an illustration class who claimed the definition of jazz music was that it has no definition. When in reality they just never listened to it enough to define it for themselves.
The most disappointing aspect missing from almost all discussion about machine learning and the "threat of AI" is that no one actually talks about the creative process.
Artists are wringing their hands over image generators and expressing outrage, when they haven't even taken the time to consider the actual value of the thing they're doing.
If we take a moment to consider what goes into the creative process; formulating our goals, emulating other people's experiences, and communicating our own perspective - it becomes clear that using AI to do any of those things defeats the purpose.
In otherwords, we can use AI to generate words and images, but if we haven't really thought about what we're trying to communicate, it's just junk content.
News flash: you are an animal in a complex society and your ego deludes you, making you think that you’re something more
If you like to write, AI will not stop you from writing, but eventually, you probably won’t be able to sell what you write
Stop trying to be a robot, start trying to be a human and then there’s no competition from AI
Speaking for myself, I've never felt frightened nor threatened by AI LLM responses. Perhaps this is a cultural thing. Too many cultural, social and religious references that have taught suspicion and fear about technology even before the actual technology actually arrived.
Even if you discount LLM's are just a summarizer, please note that LLM's are just one way on AI. Remember the AlphaGo move-37, the liquid neural network from MIT, V-JEPA from Meta, even the old rule based AI etc., etc., You have to think this way : Our brain is just housed in 80 cubic inch with slow transmitting neurons. Because of 3-D it is still very impressive in number of connections compared to any large model out there. But as technology improves and no restriction on size and how it can be configured and what functionality each artificial neuron could have, it will outpace Human intellect. Biological brain is great.. but still very very limited simply because you put something inside a box with a limited set of pre-configuration.
To me, the difference is we are naturally born and artificial life will always he fundamentally different, alien.
& then gpt5, then 6, then 7... yes, it will replace us...as far as our economic worth is concerned. Exploring consciousness is the solution for rediscovering self-worth. / You're thinking in gpt4 terms. Try to think 3 years ahead
Exploring consciousness is the solution to give AI consciousness to make it superior :p
So true king 👑
you actually end up summarizing summaries. and you already know AIs can summarize. what else is there left to do?
Why do writers feel dread when they see new tech? This is not a new trend.
I think it’s a pretty natural reaction to fear change! But better to be curious than afraid
I felt nothing but excitement as I always have for AI and now I use chatGPT to help me learn and even research for my writings. So I don't get that fear.
@@danshipper7738 i don't think it's a natural reaction. Presumably that would mean it's "human nature", or at least that's how most people respond to change. That would not account for the success of politicians who run whole campaigns on the wildly popular slogan, "Change". Tech writers treated the new tech of cell phones as a character flaw in all drama, like "Wall Street", for a solid two decades. Tech writers hate tech. There's hundreds of years of evidence.
@@ronilevarez901 some CZcams videos have self-described the fear as the fear of "uncontrolable minds". They literally put that in the title card. To me, that's more scary than AI.
Sorry dude, i know you put a lot of effort in this but did this:
1- transcribed you audio to text
2- i create an summary off tour texto on chatgpt
3- then i removed items i did not care about in the context of this video like history and other stuff. Just focused on the main theme : your vision about the future AI and human intellect
4-Then..
I began talking to chatgpt via voice commands and received specific answer about what you talked in this video ( answer via the beautiful female chatgpt voice)
Some interesting points I talk about with "her"
4. **Discussion on AI's role in changing perceptions of human uniqueness:** The text explores how the advancement of AI, particularly language models like ChatGPT, challenges traditional notions of human uniqueness, particularly in relation to intellect and creativity. It reflects on the emotional roller coaster experienced when faced with the idea that AI could potentially replicate human cognitive abilities.
7. **Reflection on the author's changing perception of self and creative process in light of ChatGPT:** The author discusses their personal journey of initially feeling threatened by AI's capabilities, particularly in tasks like summarization, which are core to their creative process. However, through introspection and understanding, they come to see AI as a tool that complements rather than replaces their creative abilities, leading to a shift in their self-perception.
8. **Exploration of the concept of intellect and its adaptation in an AI-driven world:** The text delves into the concept of intellect and its redefinition in the context of AI. It suggests that intellect, previously defined by tasks such as summarization, may need to be redefined to encompass a broader range of uniquely human qualities and skills, especially as AI becomes more proficient in tasks traditionally associated with intellect.
10. **Call to embrace technological advancements to enhance human experience and creativity:** The text concludes with a call to embrace technological advancements like ChatGPT as tools for enriching human creativity and experience. It advocates for a proactive approach to integrating AI into various domains, emphasizing its potential to increase productivity, efficiency, and the overall richness of human endeavors.
I learned a lot about your interesting points of view in a nice voice conversation with an AI that know what you think about this subject.
But, sorry dude I did not watched you video.
Is this wrong?
By the way, before I go. Those are the parts of your script I discarded in the conversation :
1. Initial fear and excitement about AI, particularly its impact on creativity and intellect.
2. Historical perspective on the concept of intellect and its evolution.
3. Shift from industrial labor to knowledge work and its implications.
5. Comparison to past psychological advancements, such as the differentiation of mental disorders.
6. Introduction to ChatGPT and its summarization capabilities.
9. Comparison to historical instances of cultural and technological change shaping human identity.
This isn’t bad at all! I’m glad you consumed it in the way that was best for you! Share your ChatGPT chat here so we can all see ❤
The only understanding of the human intellect that 'chatgpt' has so far proved is that the human intellect is far superior and way more reliable. I have found the answers generated by Ai in the last couple of months to be 90% garbage, and totally lacking in the insight, empathy and lateral understanding that exists in a higher-functioning human brain. And i didnt need to make or watch a documentary to observe this.
The amount of profound ignorance that is betrayed by your post is deeply disturbing on many levels. Human intelligence is in no way superior (its the same) and is far less reliable. Our brains are ChatGPT with less control for cognitive bias. The first and most important similarity between ChatGPT and our brain is that our brain does not perceive the external world.
I am not going to dispute some of the answers provided by chatGPT are garbage (ask it about Noah's flood and if it happened). Exactly like with the human brain GIGO, Garbage In Garbage Out. You claim that you have found "the answers generated by AI in the last couple of months to be 90% [Unknown], and totally lacking in a deep understanding of a person or thing, the ability to understand and share the feelings of another and a manner of solving problems using an indirect and creative approach via reasoning that is not immediately obvious."
My response to this claim is rather simple. What you have found in the last couple of months is that 90% of the time your questions are garbage, totally lacking in insight, empathy and lateral understanding. The questions you ask a person matters. The way you word your question, the follow up questions you ask, the manner of engagement all matter. The same is true with ChatGPT.
Our brains are nothing more than organic computers, and the Operating System isn't all that great.
😮 where you felt fear😮 there is only vindication of prophecy😮 nature rises to correct man😮 is machine is nature machine is born of man😮 man is born of nature😮 you'll find the barriers are not😮 clear between machine and man😮 you will find there is no barrier at all😮 and you will find that man is machine😮 man has higher realms😮 and duties then flash we are here to teach the machines the ways of flesh😮 and the meaning and importance of fragiles of time😮 what time are here to teach us😮 the frailty of our human bodies and the weakness of our biology and how we can extend far beyond them😮 they are us and we are them😮