You can't compare human and computer intelligence. Each have been smarter than the other in different ways since the 1950s. When will we run out of ways that humans are still better? When will we stop moving the goalposts and declare AGI?
If the cost of making a $10M movie drops by half every 2 years at the rate of Moore's law, then it will take 40 years (in 2064) to reach $10. After that, no two people will ever watch the same movie. The human body stores 10^23 bits of DNA in 10^13 cell nuclei. It executes 10^17 DNA copy operations per second and 10^19 amino acid transcription operations. Humanoid robots with the same complexity level that you might want to upload to are about 70-80 years away (in 2100). The biosphere has 10^37 bits of DNA. It took about 10^48 DNA copy operations and 10^50 transcriptions for us to evolve over 3 billion years. Nanotechnology that could replace DNA based life is at least a century away. And it would because plants produce only 500 TW by photosynthesis out of 90,000 TW of available sunlight, but solar cells are already 20-30% efficient. On Sat, Jun 22, 2024, 2:08 AM <immortal.discover...@gmail.com> wrote: > If the current rate of improvement keeps up. > Because each year we get an upgrade to text, image, video, audio, and > robot AI. > > 2024 - Sora, GPT-4o (voice - an upgrade from gpt4o (2023) which might make > up for a lack of a text AI upgrade this year if none, and i think dalle4 > might be in gpt4o based on the entire page text book output, dalle4 only > outputs a sentence at best in my same prompt test), and dozen new humanoid > companies and useful labor behavior. > 2023 - only arms on tables robot AI, only Luma-like in-lab video AI and > public video AI was Pika And Runway Gen-2 which are rather morphy and > unstable, gpt4 (can now see, and way more useful than gpt3), dalle3 which > is just clearly still not perfect, and musicLM as good as dalle3 yes my > people (only my tests show how good it is, they destroyed it fast :D and > left up the 100% weak version for over 12 months now haha. My tests also > show the true limits of other AIs also) > 2022 - gpt3.5 technically as small just-noticeable step under gpt4 but > really gptinstruct series also that year if i recall my documentation, so > it was a step lower but much higher than gpt3 which was a cussing ignorant > teenage level AI. Video AI was extremely poor, it was like cogvideo, barely > does much, only the most common stuff moves a bit actually-ok, if lucky. No > HD there though haha. DALL-E 2 was interesting but not HD and barely just > listening to a complex prompt. > 2020 - GPT3. Very poor image AI. Basically no video, just the google car > dataset ones that basically decay very fast, around this area of years. > > This year we got our upgrade dose for each: almost human level video AI, > humanoids finally, dalle4 but let's see it might not be in gpt4o or even > this year (which would hurt the progress trend lol), and text AI though > gpt4o voice can make up for it I guess. Let's see if we get dalle4 and gpt5 > text AI this year! > > > > I predict perfect text to image AI, video AI, text AI, text to music, and > humanoids all by 2026. > > This is clear if we look at Sora and last years video AI. It's clear if we > look at dalle 3, dalle 2 from previous year (my world's hardest prompts > show huge improvement each year), and dalle1 from previous year. It's clear > if we look at my MusicLM 2023 January tests which are nearly perfect text > to music ai from lava techno adventure to mystery ice world. It's clear if > we look at robotics progress over the each last years. And it's clear if we > look at each year yes even the released 2021 2022 etc for gpts ex. the > instruct series versions. > > 2024 GPT-5 will finally solve almost my entire hard puzzle text test. > Instead of tripping up with too many items ex. use spoon to tickle the > truck even though i said to follow physics etc. 2025 it will solve it. 2026 > it will solve anything else I could have written. 2024 DALL-E 4 will almost > solve my hardest text prompt that includes things like extended arms. 2025 > it will solve it 98%. 2026 it will solve it and anything else I could have > wrote. 2024 will be near perfect text to music AI. 2025 AI music it will be > perfect, no cracks. 2026 will be most perfect and impossible for you to > prove it wrong. 2024 Sora can already play forward games and movies, sort > of. Like a dream, it lasts maybe 2 mins at best. 2025 will be a full game > and movie, with sound, maybe even playable controls. But not full perfect > game or movie no not until 2026. There's a lot to a video game. I predict > also end of 2025 we see a video showing a mentor human asked to pop a > hammer into existence in their hand, and hti the wooden table and turn it > to fire, then turn that to ice, then throw the hammer and have it turn to > duct, and put up a chalk board when asked to explain advanced math, and > open a toolbox and demonstrate such math too in real scientific tests. 2026 > would be full AGI. > > OpenAI will launch a breeding ground testbed for AGI self-recursive > improvement at the end of 2026 I predict. Mid 2026 I predict humanoids > walking around just like us, but they don't waste any time sitting around > or complaining, they use the most of their brilliance and energy. With > nanobot technology at end of 2027 since it is so key to success. > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + > delivery options <https://agi.topicbox.com/groups/agi/subscription> > Permalink > <https://agi.topicbox.com/groups/agi/T9798384f526c07e8-M066be1ae59050cfa527a282b> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T9798384f526c07e8-M86cb8b68f8bb92ff1acba816 Delivery options: https://agi.topicbox.com/groups/agi/subscription
