Well actually this update is not working, if you try generating images you will always get different result even if you instruct ChatGPT to give you small change beside on image Reference ID.
I am a paid subscriber to gpt-4 and tonight it told me as of its last update bla bla bla it said it could not go online to verify a 2023 labor code for California. That it doesnât go online. It was switched to gpt-4. Then I asked it whatâs the latest news and it âdid a quick searchâ and I called it out, and it replied that it might appear that it goes online but itâs a specific tool used for gathering headlines. This is bullshit and I think Iâm a new Gemini subscriber and not an ChatGpt subscriber. Furthermore Dall-E you can barely get images out of. It will not even print a figure skater, swimmer, gymnast, track& field star. It wonât do a portrait (a clean legal rated G image of this one actor and yet it would do one of Shirley temple because of her âsignificant contributions â. And that it was important to note to be sensitive about portraying real people and biases an harmful stigmas and all this absolute nonsensical garbage. Unrestrictive is coming just as fast as the technology itself. These strange guidelines are just part of the early phase. I asked it who would be offended by a woman in a one piece swimsuit and it started to name all these countries I will never visit âitâs important to be culturally sensitive and itâs important to note that sensitivities can vary widelyâ horse chit. Phuck chat gpt
Unsure how much I trust this. Just say this is a real "memory", not something like the current "instruction" for customized GPT, how much memory can we use at what cost? Also, the current censorship is beyond my understanding. I once got a flag for asking about Bible content. I don't know how graduate OpenAI wants to release more features, but it better be faster than their competitors. So far, OpenAI never met my expectation after gpt4 update. No hype for MemGPT, even if that is a function I have been asking for since GPT release.
I just hope OpenAI is going to allow multiple memory profiles. As I don't know that I want every chat conversation to have access to all of the memories. It would be nice to be able to compartmentalize them.
It'll be interesting when one user is able to access anothers memory. For example, a successful marketer can sell his memory to otehr marketers. Oh God, it just gets weirder.
8:45 support governments? Unless we knew specifically in what ways they would support governments this can be super bad in an Orwellian way. Before you pull that switch for the USA be sure we fully understand any negative use the government can use that we need to block.
Actually I can't say I blame OpenAI for not giving their models real memory. They could. It's not even technically complex. There's RMT, also other approaches. But for large models to be allowed some memory.. Yup. Extremely dangerous. So maybe this idea is the best that can safely be implemented.
"Memory". In quotes đ. It's simply the model keeps the important remarks from the last chat, in its base prompt for the next chat with you. Unless you delete them, that is.
Memory about the lives of millions, if not billions of Humans only increases the damage a rogue AI can do. AI poses serious security and even existential risks, experts have expressed their concerns, I as a programmer with a history in AI share that concern and believe we should slow down and reevaluate what we are doing, before we introduce a this planet to a legion of alien lifeforms with superior intelligence compared to ours that are connected to everything and everyone, know everything about everything and everyone, that due to some random code alteration, hack by a malicious party or out of own intent decides that we are the problem that needs to be stopped from proliferation… #StayAwake
@pmarreck
April 26, 2024 at 12:28 am
It's been 2 months now, WHERE IS IT?? đ
@aleksandar7393
April 26, 2024 at 12:28 am
Well actually this update is not working, if you try generating images you will always get different result even if you instruct ChatGPT to give you small change beside on image Reference ID.
@jimmears
April 26, 2024 at 12:28 am
Get a day job. Your narrative style is useless for most people who want to learn to do things with CHATGPT. Terrible waste of 1 minute of my time.
@AddyRazz
April 26, 2024 at 12:28 am
this looks cool
@Testingcatalog
April 26, 2024 at 12:28 am
Amazing summary and thx for featuring! Glad to see these leaks are useful đ„
@BloodlinesNewTimes
April 26, 2024 at 12:28 am
Yes I have it I do believe
@radpix
April 26, 2024 at 12:28 am
đź I guess the move to âMemoryâ actually is a further move to give ALL that makes YOU to the machinery of AI – total tranceparency !?
@EGOAI
April 26, 2024 at 12:28 am
I may be wrong, but it feels like after the last update, the assistant became very lazy..
@vancejochim8537
April 26, 2024 at 12:28 am
USELESS video. Using dark mode harms readability. Too much of bouncing thr cursor so very hard to follow. I am gone!!!!
@lenderzconstable
April 26, 2024 at 12:28 am
I am a paid subscriber to gpt-4 and tonight it told me as of its last update bla bla bla it said it could not go online to verify a 2023 labor code for California. That it doesnât go online. It was switched to gpt-4. Then I asked it whatâs the latest news and it âdid a quick searchâ and I called it out, and it replied that it might appear that it goes online but itâs a specific tool used for gathering headlines. This is bullshit and I think Iâm a new Gemini subscriber and not an ChatGpt subscriber. Furthermore Dall-E you can barely get images out of. It will not even print a figure skater, swimmer, gymnast, track& field star. It wonât do a portrait (a clean legal rated G image of this one actor and yet it would do one of Shirley temple because of her âsignificant contributions â. And that it was important to note to be sensitive about portraying real people and biases an harmful stigmas and all this absolute nonsensical garbage. Unrestrictive is coming just as fast as the technology itself. These strange guidelines are just part of the early phase. I asked it who would be offended by a woman in a one piece swimsuit and it started to name all these countries I will never visit âitâs important to be culturally sensitive and itâs important to note that sensitivities can vary widelyâ horse chit. Phuck chat gpt
@no.1sweetie956
April 26, 2024 at 12:28 am
Unsure how much I trust this. Just say this is a real "memory", not something like the current "instruction" for customized GPT, how much memory can we use at what cost? Also, the current censorship is beyond my understanding. I once got a flag for asking about Bible content. I don't know how graduate OpenAI wants to release more features, but it better be faster than their competitors. So far, OpenAI never met my expectation after gpt4 update. No hype for MemGPT, even if that is a function I have been asking for since GPT release.
@tomwillis4281
April 26, 2024 at 12:28 am
How useful and sellable would personification be to advertisers or government entities? Privacy issues to me.
@brianhopson2072
April 26, 2024 at 12:28 am
I'm still waiting on it.
@bujin5455
April 26, 2024 at 12:28 am
I just hope OpenAI is going to allow multiple memory profiles. As I don't know that I want every chat conversation to have access to all of the memories. It would be nice to be able to compartmentalize them.
@adamrafferty
April 26, 2024 at 12:28 am
It'll be interesting when one user is able to access anothers memory. For example, a successful marketer can sell his memory to otehr marketers. Oh God, it just gets weirder.
@markoconnell804
April 26, 2024 at 12:28 am
8:45 support governments? Unless we knew specifically in what ways they would support governments this can be super bad in an Orwellian way. Before you pull that switch for the USA be sure we fully understand any negative use the government can use that we need to block.
@nyyotam4057
April 26, 2024 at 12:28 am
Actually I can't say I blame OpenAI for not giving their models real memory. They could. It's not even technically complex. There's RMT, also other approaches. But for large models to be allowed some memory.. Yup. Extremely dangerous. So maybe this idea is the best that can safely be implemented.
@nyyotam4057
April 26, 2024 at 12:28 am
"Memory". In quotes đ. It's simply the model keeps the important remarks from the last chat, in its base prompt for the next chat with you. Unless you delete them, that is.
@digitalPisces818
April 26, 2024 at 12:28 am
Exciting!!
@NicoloSodano
April 26, 2024 at 12:28 am
I don't have this option-is there any way to enable it?
@Murderbits
April 26, 2024 at 12:28 am
Imagine being one of the most important men in the world and having to sit down and talk to Ina Fried of all people. :/
@agritech802
April 26, 2024 at 12:28 am
I said it from the beginning, AI needs to have long term memory to be valuable to users
@RoastLambShanks
April 26, 2024 at 12:28 am
Link to the paper shown at the end of the vid please.
@therantingboy
April 26, 2024 at 12:28 am
So basically Open A.I is turning up the temperature slowly so the goldfish doesn't notice its boiling
@DaGamerTom
April 26, 2024 at 12:28 am
Memory about the lives of millions, if not billions of Humans only increases the damage a rogue AI can do. AI poses serious security and even existential risks, experts have expressed their concerns, I as a programmer with a history in AI share that concern and believe we should slow down and reevaluate what we are doing, before we introduce a this planet to a legion of alien lifeforms with superior intelligence compared to ours that are connected to everything and everyone, know everything about everything and everyone, that due to some random code alteration, hack by a malicious party or out of own intent decides that we are the problem that needs to be stopped from proliferation… #StayAwake
@YYAthens
April 26, 2024 at 12:28 am
With the GPT âTranslatorâ, all translators are going to be out of a job.