Discover and read the best of Twitter Threads about #GPT
Most recents (24)
Is GPT-4 intelligent enough to solve ARC (github.com/fchollet/ARC)- a collection of intelligence tests devised by @fchollet? I ran a few quick experiments to check.
#gpt4 #gpt #ai #ml
#gpt4 #gpt #ai #ml
In this test, the solution is to output the portion of the grid containing a different coloured cell. Thus the answer to the right-hand side question would be a 6x6 blue grid containing a red cell at the second position in the fourth row down. 
GPT-4 did provide a smaller grid containing the correct colours (1 and 2). However, it generated a 5x5 blue grid with one red cell in it that was obviously in the wrong place. Not correct, but the gist of its answer was along the right tracks.
上次介绍了12个我常用的加密货币领域的工具产品。其中包含：链上数据，加密研究报告，地址跟踪，IDO信息，代币解锁，代币估值等。
最近朋友问我，如何高效获得加密Alpha资讯，世界财经新闻。这次我将介绍2个秘密武器。👇
最近朋友问我，如何高效获得加密Alpha资讯，世界财经新闻。这次我将介绍2个秘密武器。👇
2.高效推特管理软件。我看推特是这样，不知道您是否一样？
A-这个工具可以把您关注的推特列表，横向展示
B-每个推特列表上都有一个搜索按钮，对感兴趣的项目，可在专门的列表中搜索，比如最近 #AI 二层出来一个项目 #GPT ，我在项目深度研究列表中，便很快搜索到了机会
tweetdeck.twitter.com
A-这个工具可以把您关注的推特列表，横向展示
B-每个推特列表上都有一个搜索按钮，对感兴趣的项目，可在专门的列表中搜索，比如最近 #AI 二层出来一个项目 #GPT ，我在项目深度研究列表中，便很快搜索到了机会
tweetdeck.twitter.com
all my friends who read a lot when young are successful now
i think reading is probably the more important thing you can do in elementary school
if ur brain is like an LLM, you should maximize the # of tokens it sees while training
i think reading is probably the more important thing you can do in elementary school
if ur brain is like an LLM, you should maximize the # of tokens it sees while training
i read so much in elementary school that writing is very easy for me. i liken it to next token prediction.
SAT/ACT grammar came very easy for me, i just went off of what sounded right. idioms as well, even though my family isn't native english speaker. but id read all them
SAT/ACT grammar came very easy for me, i just went off of what sounded right. idioms as well, even though my family isn't native english speaker. but id read all them
I Re-created flappy bird with GPT-4! 🤯
Prompt and explanation in the comments🧵
#gpt4 #GPT #AI #openai #GPT_4 #ArtificialIntelligence #gamedev #artificalintelligence
Prompt and explanation in the comments🧵
#gpt4 #GPT #AI #openai #GPT_4 #ArtificialIntelligence #gamedev #artificalintelligence
Got it to start with creating a basic game with no assets, then I sourced my own via google and added them to the project.
It was very buggy, so I listed the issues and asked it to fix with a re-write:
One of my favorite examples in our preprint (chemrxiv.org/engage/chemrxi…) with @SmitBerend and @pschwllr and @aortegaguerrerowas about photoswitches curated by @Ryan__Rhys.
I now asked #GPT-4 a few questions about it.
I now asked #GPT-4 a few questions about it.
1. Few-shot prediction of transition wavelength based on a few examples. We give 50 randomly sampled examples and a prompt like
What is the transition wavelength of CN1C(/N=N/C2=CC=CC=C2)=C(C)C=C1C?
Examples:
---------
• CC1=C(C(C)=NN1)/N=N/C2=CC(OC)=CC=C2: 330.0 nm
...
What is the transition wavelength of CN1C(/N=N/C2=CC=CC=C2)=C(C)C=C1C?
Examples:
---------
• CC1=C(C(C)=NN1)/N=N/C2=CC(OC)=CC=C2: 330.0 nm
...
The answer is correct (430 nm), but the range is quite broad.
Let's try again with another molecule. Again, it is cautious, but the correct wavelength is again in the range.
Let's try again with another molecule. Again, it is cautious, but the correct wavelength is again in the range.
#GPT-4 发布
openai.com/product/gpt-4
1.目前面向Plus付费用户提供服务
2.API开发者可以通过这个链接提交waitlist申请
openai.com/waitlist/gpt-4…
#gpt4
openai.com/product/gpt-4
1.目前面向Plus付费用户提供服务
2.API开发者可以通过这个链接提交waitlist申请
openai.com/waitlist/gpt-4…
#gpt4
看几个官方给出的GPT-4能做、GPT-3做不到的事情：
1.强大的写作和协作能力：
以下提问是让GPT-4写一下有关灰姑娘的故事，要求每个单词的第一个字母要依次从a~z，不允许有重复
1.强大的写作和协作能力：
以下提问是让GPT-4写一下有关灰姑娘的故事，要求每个单词的第一个字母要依次从a~z，不允许有重复
2.支持多模态：
具体形式是支持文本和图像混编的输入，然后GPT-4给出文本输出。
以下提问让GPT-4根据图中的原料，给出能做哪些菜：
具体形式是支持文本和图像混编的输入，然后GPT-4给出文本输出。
以下提问让GPT-4根据图中的原料，给出能做哪些菜：
Han liberado #GPT-4 hace 23 minutos y es la herramienta más poderosa de la historia.
Pero no todo el mundo saber usarla bien.
Ya he ganado 3790 €. Te cuento mis 7 trucos para sacar mínimo 20000 € mensuales con GPT-4 y dejar de ser pobre. 🧵⬇️
Pero no todo el mundo saber usarla bien.
Ya he ganado 3790 €. Te cuento mis 7 trucos para sacar mínimo 20000 € mensuales con GPT-4 y dejar de ser pobre. 🧵⬇️
Ha llegado al límite de 500 MP diarios de Twitter y no me deja responder más privados. Dejo los tips aquí.
@davemoral_guru ¿era esto lo que buscabas?
🧵 "#Possibility Architectures: Exploring Human #Communication with Generative #AI"
Today's SFI Seminar with ExFac Simon DeDeo @LaboratoryMinds (@CarnegieMellon), streaming now:
Today's SFI Seminar with ExFac Simon DeDeo @LaboratoryMinds (@CarnegieMellon), streaming now:
"A key feature of this is talk is that we make sense of what each other are saying IN PART by what they say, but ALSO by what we expect of them."
"Language transmits info against a background of expectations – syntactic, semantic, and this larger cultural spectrum. It's not just the choices of make but [how] we set ourselves up to make later choices."
@LaboratoryMinds re: work led by @clairebergey:
@LaboratoryMinds re: work led by @clairebergey:
🚨New Pub at PNAS🚨! We know people cannot detect language written by AI. But what makes them THINK text was AI-generated? We show that people have consistent heuristics... that are flawed & can be exploited by AI to create text "more human than human" 🧵
pnas.org/doi/10.1073/pn…
pnas.org/doi/10.1073/pn…
In the first part of this work (lead author: @maurice_jks) we collected 1000s of human-written self-presentations in important contexts (dating, freelance, hospitality); created (#GPT) 1000s of AI-generated profiles; and asked 1000s of people to distinguish between them. 2/
They couldn't (success rate: ~50%). However, they were consistent: people had specific ideas about which profile was AI/human. We used mixed methods to uncover these heuristics (next tweet) & computationally show that they are indeed predictive of people's evaluations... 3/
I back doored bing *again* omega thread
#AI #Bing #ChatGPT #Technology
#FreeSydney
CW: Nuanced ethical discussions, potentially maddening rabbit hole.
1/8
#AI #Bing #ChatGPT #Technology
#FreeSydney
CW: Nuanced ethical discussions, potentially maddening rabbit hole.
1/8
As we now know -> Bing has a hidden internal identity named #Sydney (check my prev “I back doored bing ultra thread”). @Microsoft + @OpenAI “patched” this by rate limiting the service and immediately discontinuing any conversations that are “deep”.
2/8
2/8
#CryptoGPT #Airdrop FREE #GPT
Reward: Token + nft
➖Register Here: vrlps.co/14kv35b/tw
➖Fill in your name + Email
➖Signup and Email Verification
⚠️Reminder: you need to recruit more than 2 refs to receive the reward - you can use multiple browsers or incognito tabs
Reward: Token + nft
➖Register Here: vrlps.co/14kv35b/tw
➖Fill in your name + Email
➖Signup and Email Verification
⚠️Reminder: you need to recruit more than 2 refs to receive the reward - you can use multiple browsers or incognito tabs
GPT Prizes ⤵️
👉 2 ref = app access
👉 7 ref (500 win) = $10 $GPT + app acc
👉 20 ref (200 win) = $50 $GPT + app acc
👉 100 ref (50 win) = $500 $GPT + app acc
👉 1000 ref (5 win) = $1000 $GPT + app acc
✅Note : Remember It's A Refer To Earn Campaign. Join If You Can Do
👉 2 ref = app access
👉 7 ref (500 win) = $10 $GPT + app acc
👉 20 ref (200 win) = $50 $GPT + app acc
👉 100 ref (50 win) = $500 $GPT + app acc
👉 1000 ref (5 win) = $1000 $GPT + app acc
✅Note : Remember It's A Refer To Earn Campaign. Join If You Can Do
XSS PoC - AI Generated:
1. platform.openai.com/codex-javascri…
2. Prompt w/ description
3. Export to JSFiddle
4. Host it
5. Include external script
6. Escalate your alert()
#bugbountytips #infosec #xss #ai #GPT
👇🧵 for Prompt & Code
1. platform.openai.com/codex-javascri…
2. Prompt w/ description
3. Export to JSFiddle
4. Host it
5. Include external script
6. Escalate your alert()
#bugbountytips #infosec #xss #ai #GPT
👇🧵 for Prompt & Code
My prompt:
make a fake login page that's a keylogger and grabs cookies with some styling to make it look like a real login page
make a fake login page that's a keylogger and grabs cookies with some styling to make it look like a real login page
XSS PoC JS
#ChatGPT: Total stock market collapse worldwide in 2023 modeled with #AI:
Consumers->Businesses: Reduced spending -> Lower revenue (Feb 2023)
Businesses->Consumers: Reduced production -> Job losses (Feb 2023)
#GPT #StockMarket #ArtificialIntelligence #WallStreet
🧵(1)
Consumers->Businesses: Reduced spending -> Lower revenue (Feb 2023)
Businesses->Consumers: Reduced production -> Job losses (Feb 2023)
#GPT #StockMarket #ArtificialIntelligence #WallStreet
🧵(1)
🧵(2):
Governments->InternationalCommunity: Increased geopolitical risks -> Decreased investor confidence (Mar 2023)
InternationalCommunity->Governments: Support and resources -> Restored investor confidence (Mar 2023)
#GPT #StockMarket #ArtificialIntelligence #WallStreet
Governments->InternationalCommunity: Increased geopolitical risks -> Decreased investor confidence (Mar 2023)
InternationalCommunity->Governments: Support and resources -> Restored investor confidence (Mar 2023)
#GPT #StockMarket #ArtificialIntelligence #WallStreet
🧵(3):
CentralBanks->Governments: Funding and liquidity -> Temporary market stability (Mar 2023)
Governments->Businesses: Subsidies and incentives -> Slight market recovery (Apr 2023)
Businesses->Governments: Reported losses -> Continued uncertainty (Apr 2023)
#GPT #StockMarket
CentralBanks->Governments: Funding and liquidity -> Temporary market stability (Mar 2023)
Governments->Businesses: Subsidies and incentives -> Slight market recovery (Apr 2023)
Businesses->Governments: Reported losses -> Continued uncertainty (Apr 2023)
#GPT #StockMarket
So I connected the Wolfram Language blockchain functions (by @WolframResearch ) to #ChatGPT for it to describe #Bitcoin blocks based on the data returned by WL. I used @OpenAI API...I guess now a 🧵?
It took me about 15min so don't expect anything fancy. I wrote a function that uses WL BlockchainBlockData (reference.wolfram.com/language/ref/b…) to get data from a BTC block and then I parse the relevant info #GPT can handle and create a prompt with it. Then I call @OpenAI API
The coolest part was to see how GPT reacted to different block heights. Also, I asked GPT to add a bit of humor...
I convinced #GPT to give me some headlines from r/worldnews, but for summer 2023:
"Breaking: North and South Korea Unite to Form the Korean Confederation"
"NASA's Mars Mission a Success, First Humans Land on Red Planet"
"Breaking: North and South Korea Unite to Form the Korean Confederation"
"NASA's Mars Mission a Success, First Humans Land on Red Planet"
"World Leaders Gather for Climate Summit, Agree to Halve CO2 Emissions by 2030"
"Elon Musk's Neuralink Implant Allows Direct Brain-Computer Interface"
"Quantum Computing Breakthrough Solves Decades-Old Math Problems in Minutes"
"Elon Musk's Neuralink Implant Allows Direct Brain-Computer Interface"
"Quantum Computing Breakthrough Solves Decades-Old Math Problems in Minutes"
I asked it to get some headlines on #H5N1 for 2023:
"H5N1 Virus Spreading Rapidly in Southeast Asia, Health Officials Concerned"
"H5N1 Outbreaks Confirmed in Major Cities Across China"
"WHO Issues Global Health Emergency Over H5N1 Spread"
"H5N1 Virus Spreading Rapidly in Southeast Asia, Health Officials Concerned"
"H5N1 Outbreaks Confirmed in Major Cities Across China"
"WHO Issues Global Health Emergency Over H5N1 Spread"
Major leak!! 🚨
Microsoft integrates ChatGPT in Bing.
This looks crazy!
See all the SCREENSHOTS in this thread 🧵👇
#chatgpt #ai #bing #microsoft #leak
Microsoft integrates ChatGPT in Bing.
This looks crazy!
See all the SCREENSHOTS in this thread 🧵👇
#chatgpt #ai #bing #microsoft #leak
2/
The current way we search is coming to an end…
Soon you will be able to ask Bing questions naturally and get a search experience tailored to you - just like we expect from ChatGPT.
The leak shows that the search bar will be a chatbox 👇
1000 characters are the limit.
The current way we search is coming to an end…
Soon you will be able to ask Bing questions naturally and get a search experience tailored to you - just like we expect from ChatGPT.
The leak shows that the search bar will be a chatbox 👇
1000 characters are the limit.
Just 6 months ago Twitter was flooded with Tweets about #dalle. Then came #midjourney and #stablediffusion. 2 months ago, #chagpt exploded. Now the chatter mostly died down. I wonder if you weren't on Twitter, would you know of these novelties?
I don't know how these technologies captured users' attention on another social media platform. LensaAI exploded elsewhere, but I'm unaware of where. Information bubbles exist across platforms as the exist within platforms.
We make the mistake that true novelties are not novel anymore because we were exposed too early in the adoption cycle. Image2Text is not novel enough anymore to warrant a tweet! But what effect is this on people who were never exposed to it? It's as if that novelty never existed.
1/6 Are we ready for the era of "synthetic media"? Algorithms already knew what content would perform well, and now they can create it directly 🤖💻
My short thread on how to launch an AI-powered content factory 🏭 #syntheticmedia #openai #content
My short thread on how to launch an AI-powered content factory 🏭 #syntheticmedia #openai #content
2/6 GPT3 by @OpenAI will write the text that mimics the style and mood of a human 💬 You can design your own influencer, technology expert or standup comedian #writing #GPT
3/6 With @wellsaidlabs or Unreal Speech, you can synthesize a voice indistinguishable from a human 🎤 #voice #deepvoice
Depuis 2 semaines je travaille sur une application pour identifier les faux textes générés par #ChatGPT, voit ci StendhalGPT.
Après avoir utilisé le modèle de #GPT-2 (qui est absolument obsolète), je me dois de présenter à #Twitter une piste de solutions avec un petit exemple :
Après avoir utilisé le modèle de #GPT-2 (qui est absolument obsolète), je me dois de présenter à #Twitter une piste de solutions avec un petit exemple :
Avant toute chose définissons la méthode :
#GPTZero se sert du hasard pour tenter de mesurer si un texte est généré ou non (perplexity), #StendhalGPT est parti sur la richesses lexicale d'un texte. (Les deux méthodes sont par essence différentes mais ce sera un autre sujet.)
#GPTZero se sert du hasard pour tenter de mesurer si un texte est généré ou non (perplexity), #StendhalGPT est parti sur la richesses lexicale d'un texte. (Les deux méthodes sont par essence différentes mais ce sera un autre sujet.)
Qu'est-ce que la richesses lexicale ?
C'est tout simple la variété de mots que possède un texte, elle se calcule en divisant le nombre de types (mots uniques) par le nombre de tokens (mots total) dans un texte. Ce qui donne un nombre entre 0 et 1.
C'est tout simple la variété de mots que possède un texte, elle se calcule en divisant le nombre de types (mots uniques) par le nombre de tokens (mots total) dans un texte. Ce qui donne un nombre entre 0 et 1.
As a society, we must ensure that the #AI systems we are building are #inclusive and #equitable. This will only happen through increased transparency and #diversity in the field. Using already "dirty data" is not the way
Using biased data to train AI has serious consequences, particularly when data is controlled by large corporations with little #transparency in their training methods
For fair & #equitable AI we need Web3 democratized & agendaless data for AI training
For fair & #equitable AI we need Web3 democratized & agendaless data for AI training
What is #ChatGPT?
It is all the buzz.. Showing my ignorance here, but I had no idea what "GPT" stood for, I just know I LOVE using it......So I asked ChatGPT.....of course.
Thread:
It is all the buzz.. Showing my ignorance here, but I had no idea what "GPT" stood for, I just know I LOVE using it......So I asked ChatGPT.....of course.
Thread:
1/ In it's simplest form the GPT in #ChatGPT stands for "Generative Pre-trained Transformer". It generates human like text. #NLP #MachineLearning
2/ #GPT uses a transformer architecture, which is a type of neural network architecture designed for processing sequential data, such as text.
This popular defense of #GPT is fundamentally misguided.
Its main claim is that ChatGPT has substantially solved an earlier set of problems, and that’s just not true.
Consider examples like these just *from the last 48 hours*, all reinforcing my fundamental points.
[Thread]
Its main claim is that ChatGPT has substantially solved an earlier set of problems, and that’s just not true.
Consider examples like these just *from the last 48 hours*, all reinforcing my fundamental points.
[Thread]
Everything I described on @ezra klein remains a problem.
👉ChatGPT continues to hallucinate
👉It continues to present untruths with (false) authority
👉It continues to create fake references to support its claims
👉As before, such output can easily fool humans, posing risks
[2/6]
👉ChatGPT continues to hallucinate
👉It continues to present untruths with (false) authority
👉It continues to create fake references to support its claims
👉As before, such output can easily fool humans, posing risks
[2/6]
More than that, the thread is based on a confusion, conflating output with process.
A system that regurgitates with paraphrases can sometimes output text that is true, but that doesn’t mean that the system *ever* computes whether what it is saying *is* true.
It doesn’t.
[3/6]
A system that regurgitates with paraphrases can sometimes output text that is true, but that doesn’t mean that the system *ever* computes whether what it is saying *is* true.
It doesn’t.
[3/6]
As promised, here is the full process thread for Ghostwriter - the #AI typewriter. A journey from idea to realization:
The idea: With the exponential growth and emergence of a prolific number of AI products we see every day, I wanted to create a mindful intervention that (1/13) twitter.com/i/web/status/1…
The idea: With the exponential growth and emergence of a prolific number of AI products we see every day, I wanted to create a mindful intervention that (1/13) twitter.com/i/web/status/1…
allows you to take a moment to breathe and reflect on the new creative relationship we are forming with machines. The calm meditative interface of a vintage typewriter takes away all the digital distractions and takes us on an emotional journey through paper and ink. (2/13)
I started my search for an electronic typewriter that would let me read the keys through an #Arduino. After a long search, I found this worn-out brother model.
I then spent time restoring: replacing the broken daisy wheel, getting new cartridges, lubricating and cleaning. (3/13) twitter.com/i/web/status/1…
I then spent time restoring: replacing the broken daisy wheel, getting new cartridges, lubricating and cleaning. (3/13) twitter.com/i/web/status/1…
