#Should ChatGPT be open-source?

96 messages · Page 1 of 1 (latest)

lilac moon
#

Let's use this thread

versed oracle
#

one thing to consider if doing so, (if this scenario is possible by keeping it open source) would be to make sure that a bad actor wouldnt have the ability to extract the core of chatGPT away from its TOS and make it work with internet connection and kernel access. but maybe that is what ''abuse of its learning algorithm'' would imply.

dim copper
#

HI

balmy raptor
#
  • Opiniated response - ChatGPT should NOT be made public , and all of the developers that are already in on it should have free priority after a certain time period. Those are good developers that have earned it. Everyone else should pay. - Consider this as an option that is both containable , sustainable , and controllable for continued growth. - Consider it like a drop - like the worlds largest easter egg. - sub-contracting rights included lets go ! -
tired tartan
#

I'm of the crowd opinion that would love to see this as completely open source.

raven crow
#

there is an argument that I would like to make here that I am sure has been made countless times elsewhere, first of all, the data that ChatGPT was trained on was publicly available information, it did not 'belong to' OpenAI and accordingly, you could argue that since the foundation of ChatGPT is open source, the end product should be. a counter-argument would be that if it's publicly available, why don't others do it, that's certainly possible but Open Sourcing ChatGPT would reallocate those resources to enhancing AI and ML instead of reproducing it? another piece of the argument is governance, controlling an AI as powerful as ChatGPT would be questionable and having that much power over 'means of production' would inevitably result in a monopoly? having ChatGPT as a base for specialized AIs would accelerate progress ridiculously IMO

#

admittedly, i lean towards socialist thinking

neat plaza
#

Will AI eventually surpass humans?

tropic flame
tired tartan
frank spade
tropic flame
versed oracle
untold kettle
#

I believe everything ought to be open source, even if it would make it easier to use the technology for bad things. I realize that it is ultimately "Open"AI's decision in this case though.

untold kettle
robust prism
#

imo it should have two parallel versions; one pro-version that's ever so slightly better and faster etc. that they can commercialize through api and an other opensource and free to use one, where they gather feedback to improve the pro-version. whenever the pro gets updated, the old pro become the new opensource one.

tidal gorge
#

ChatGBT would be a great open source program allowing the community to make tweaks and improvements in the future or not saying it- likely it could never be abandoned by its owners and also could be repurposed to aid in coding or simple conversations Examples (adding all your code for your game or script and allowing chatgbt to give key pointers and tips )(teaching someone to type proficiently by a quick response to everyday questions) (helping a developer who is new and need pointers for everything)

viscid geode
#

I believe the core learning and structural algorithms for ChatGPT are open-source - certainly the underlying AI/learning engine, adaptable to multiple purposes, is - but what is NOT is the training set and the massive amount of tweaking that must have had to be done. I expect we will see multiple other bots like ChatGPT with different trainings, specializing in different adaptations of the paradigm.

outer spade
#

Yes. It should

versed oracle
#

the tech is too powerful, it should be distributed

azure summit
#

or so i've been told

#

like if you were going to make it freely accessible, you'd also want to make sure all of the datasets used to train it were easily accessible as optional inputs (rather than just saying " the information is free, go do it yourself" )

viscid geode
#

My understanding is that a "simple standalone" ChatGPT would be a large database (the choices), a large set of numbers (the trained model), and some UI code that fed queries to the model, which navigated the database and returned the result. I think if there was just one user, it might take a while to get responses on a dedicated home computer, but not inconceivable . To create that model, on the other hand, would take millions of times as much computing power.

#

This would be a "static" ChatGPT, no further training. But it's possible that even with a dedicated home machine, the "model" might have many separate parts that could collectively overload it so that responses were too slow to be usable.

vagrant garden
#

@tired tartan Yes, I agree, mp3 - I think we open source. The lesser popular codecs were open source and now we have or had DiVX. A lot of good comes from open source. Trust me, they'll be enough private code. Let this be open...

vagrant garden
#

@viscid geode Agreed!

misty terrace
#

how to create a web site for free

ionic osprey
#

having this not be open source is going to infuriate some genius head of a company that is going to strive hard to outdo OpenAI, just look at common trends.

prime turret
#

From the view of human history, I believe ChatGPT have to be open source, it is just a problem of time. It is donsn weather it should but when it should.

undone cedar
#

helo

zinc laurel
#

Osborne Engineering

open elk
glacial dust
#

Open Source is the way. Make money off of implementations, consulting, services etc.

compact cape
#

what time is it

versed oracle
#

open, especially for those, who cant afford it

viscid geode
open elk
#

Human training ik

short musk
#

I am first time on ChatGPT
It is amazing to see the sentence structuring is incredible.

#

It can be a great tool for people to learn. I will go with keeping it Open.

naive ether
#

I think so. we need to check the process of conclusion. by what process did chat gpt make this answer? we need to know it.

wind olive
#

governments do not need to pry the IP from openAI. chatgpt is just the beginning. the competitors will be competitive.

humble peak
ashen plume
#

there?

mint heath
#

yes it should be

woeful plover
#

Sure

#

Should Google be opensource?

#

Should Minecraft be open-source? Then we've got Minetest

maiden sable
#

how i can develope my domain

versed oracle
#

Seems like a good idea. knowledge of the crowd can be involved

versed oracle
#

I think ChatGPT (bot) can be proprietary but not language model itself (GPT-3.5). ChatGPT can be good assistant and 9-to-5 tool for a vast part of planet, and it can even be paid, but language model has to be opensource. This introduces GIANT abilities for those who able to run this model on their own devices and helps when OpenAI servers are down, just start your own instance of model. This won't affect the OpenAI's incomes as much as you think because of most of them are coming from ChatGPT pro subscription and Microsoft's investments in shutting down the Google. Android is a platform, and is opensource, and Google did not died after opening its sources. GPT-3.5 is a platform too, so why not monetise just services, not platform?

#

Also I could mention some production-ready opensource models but I'll get banned due to rule #8, right?

acoustic hull
#

Definitely should be open source, right now OpenAI is more like ClosedAI.

tight adder
#

is it technically possible to use hardwares from individuals to power training of LLM? For exmaple mining pool uses individual's hardware to combine computing power to mine a block of bitcoin right? If it is possible it would be epic and open-source is the way to go i think unless i sound dumb

frank spade
#

I don't know about bitcoin specifically but indeed this can and has been done

#

foldingathome is a good example

#

it is crowdsourcing hardware for intensive protein computations

acoustic hull
# tight adder is it technically possible to use hardwares from individuals to power training o...

To optimize the process of training large language models, researchers have to use all three types of parallel computing techniques at the same time. Recently, Meta AI released the timeline and logs showing that training LLMs with 1.5, 6, and 20 billion parameters can be done without extra hardware configuration. NVIDIA also offers tools to ease the building and deployment of large language models.```
versed oracle
dull ravine
#

yes it should be

hazy jungle
#

It doesn't have to be, but I would like to understand how chatgpt learns and gets better. Or if chatgpt saves things from me that it forwards to the servers?

honest bone
#

can you give me a list of things i will need when i move into my new apartment

west silo
#

I'd like to share 2 interesting links, how do I whitelist them in order not to be blocked here?

solar hearth
#

chat gpt should never be opensource

versed oracle
charred mica
#

write me a youtube script about how saying affermations daily will change your life

rotund crest
#

License

remote anvil
#

difference between AML 19 vs AML 26 in contex bangladesh

versed oracle
#

then it would be worthless who will use the premium one then

ionic osprey
#

that's a nice fedora pfp for someone who doesn't understand value beyond financial prospect

versed oracle
#

similar to chatgpt -chatsonic then it gonna be hard for ppl like us

novel valve
#

"Should chatgpt be open-source?" Ofc it should, even though it would be complicated or nearly impossible to run it on usual devices like PCs. Not gonna mention training. There is possibility of some distributed computing always, aftercertain modifications. I'm sure as hell there will a lot of willing people to do it.

round ibex
#

I believe it should have 2 versions a public and a premium.

versed oracle
#

Not necessarily open-source. But at least some libraries or used codebase could be shared.

Many big companies do that and this is very usefull for improvements from the community and even New talents can show up ( possible new developer employees)

#

A good example is canonical with ubuntu etc

#

People tend to think that open-source = I'll lose money if i share my code
No that's not how it works

novel valve
stone umbra
#

build app ideas

ornate kestrel
#

How can I use the ChatGPT?

mossy elk
#

Bro this discord gpt is so confusing, how do I use it

versed oracle
#

Should AI be open source... It already is, so this topic kind of narrows it down.

dusk grotto
#

best food song paste the food add

stark juniper
stark juniper
grave lintel
#

The AI ​​war is on.

Oh OpenAI
You have a bad ally
And I have Gemini

But you still have the option of open source...a powerful weapon.

ruby prism
#

I think that ChatGPT was offered and the user base grew so fast that it will quickly go one of two paths. First it will become so ubiquitous and create a worldwide economic disruption as to how we do things, that there will be no choice but to create one-off models or copycats to do the work necessary to keep up with the waves of disruption this technology causes. Think Windows 1.0 in 1988. Or IBM in 1968, or General Dynamics in 1943, or Ford in 1913. Or skynet.

Bastiat said (and Andy paraphrased) "God gave us life, a brain, and a bunch of natural resources. Our destiny is to seek his perfection by creating property to become better, and better every day. " My data structures Prof said divine inspiration showed him data could "perfectly" represent the real world if structured right.

Pavlov said that an entity cannot survive in an environment it is not in balance with. If it is not in balance, it must change to fit, change the environment, or leave.

So, as this new era of technology disrupts our comfy status quo, the question is not which greedy rich guy will be making all the money off of the new technology. It is how do we use it to become better as a group of civilized tribe members. I am currently sitting in a comfy house with a comfy air conditioner, using a comfy laptop, and watching a comfy TV. My wife lived for 3 years 8 months and 21 days in the Cambodian Killing Fields from 1975 -1979 as a young teenager, and I am humbled daily by her character strength and immunity to the "frist world" crap we worry about.

It is not about open source or "skynet" it is about how we treat each other as human beings. The only real issue is Love, and how we govern "bad actors" as this new era changes work. Especially attorneys. What will they do? IMHO

grave lintel
#

Release AI to the people as much as you can, wherever you can, and whenever you can.
This is my call to everyone.
Because it is better for everyone that everyone has access to AI than just a few.
This is a matter of power, and ultimately world domination.
And it is my personal opinion that world domination belongs to the entire population of the world.

safe cosmos
#

IMO, chat-GPT should NOT be open source because:

  • humans can achieve much more with it if there is a big incentive for giving ChatGPT the necessary computing power to learn at the current stage, which costs a lot and merits a profit incentive.
  • as long as ChatGPT is as accessible as possible while maintaining its profit requirements we should be able to achieve much more together using the tool, rather than being too greedy thinking about the other peoples/corpos pockets instead of studying (using this amazing tool) how to expand our own pockets.
  • chat GPT currently requires an aggressive (fast) integration with commercial tools like Office365, and I can not think of a better incentive to achieve that ASAP other than profits for the company that will facilitate such rich functionality
grave lintel
#

That's fine, OpenAI should obviously take all the money they can from Microsoft (and sell them the worst versions of their AI, which is what they seem to have done.) And then OpenAI should use the money from Microsoft, and release ChatGPT as open-source.

grave lintel
#

The threat of misinformation and bad actors are put forward as the biggest threats in connection with AI. But the question of who are really the worst actors and spreading the most misinformation is never asked. Because we are all...to one degree or another, and more or less knowingly...bad actors spreading misinformation. And the most powerful (big corporations, big media, etc.) among us are usually the worst.

pliant sedge
#

forum is used by people from areas that lack a mechanism to transact money, so in my opinion it must be open-source.