Discover and read the best of Twitter Threads about #GPT4
Most recents (24)
🔥1/8
Introducing "🤖 Task-driven Autonomous Agent"
An agent that leverages @OpenAI's GPT-4, @pinecone vector search, and @LangChainAI framework to autonomously create and perform tasks based on an objective.
"Paper": yoheinakajima.com/task-driven-au…
[More 🔽]
Introducing "🤖 Task-driven Autonomous Agent"
An agent that leverages @OpenAI's GPT-4, @pinecone vector search, and @LangChainAI framework to autonomously create and perform tasks based on an objective.
"Paper": yoheinakajima.com/task-driven-au…
[More 🔽]
🚀2/8 The system can complete tasks, generate new tasks based on results, and prioritize tasks in real-time. It demonstrates the potential of AI-powered language models to autonomously perform tasks within various constraints and contexts.
💡3/8 The autonomous agent uses GPT-4 for task completion, Pinecone for efficient search and storage of task-related data, and the LangChain framework to enhance decision-making processes. #GPT4 #Pinecone #LangChain
What dates📅 can @OpenAI, @AnthropicAI, @CohereAI models reliably answer questions for?🔭
I binary-search through "future" Wiki events to find out. Results ❌🟰❌documentation:
#GPT4 ➡️~Dec 19 ('21)
#ChatGPT ➡️~Oct 24
Claude v1.2➡️~Oct 10
Cohere XL Nightly➡️~Apr 24 ('22)
1/🧵
I binary-search through "future" Wiki events to find out. Results ❌🟰❌documentation:
#GPT4 ➡️~Dec 19 ('21)
#ChatGPT ➡️~Oct 24
Claude v1.2➡️~Oct 10
Cohere XL Nightly➡️~Apr 24 ('22)
1/🧵
GPT4 says it is trained up to Sept 2021.
I found it correctly answers unknowable events in Oct, Nov, and even Dec 11th & 19th.
In late Dec it begins to abstain.
2/
I found it correctly answers unknowable events in Oct, Nov, and even Dec 11th & 19th.
In late Dec it begins to abstain.
2/
Interestingly, GPT 3.5 "Default" answers correctly only until ~Oct 24, 2021, but GPT 3.5 "Legacy" answers correctly until ~Oct 31, 2021 then begins hallucinating false answers or abstaining in Nov.
Perhaps this is due to finetuning rather than pretraining data?
3/
Perhaps this is due to finetuning rather than pretraining data?
3/
🤖 My LLM Week 1 & 2 🤖
#gpt4 was only publicly released 14 days ago
Yet, the LLM world is moving really fast, and people are noticing that it's an "iPhone moment"
But if you haven't been able to play catch up, here are the 3 top topics that caught my attention 👇
#gpt4 was only publicly released 14 days ago
Yet, the LLM world is moving really fast, and people are noticing that it's an "iPhone moment"
But if you haven't been able to play catch up, here are the 3 top topics that caught my attention 👇
1/ The paper on jobs impacted by #gpt dropped
Must join #ChatGPT Learning/Jam use @github CTO #Hack nat.dev Workshop (US) @Meetup w/ @ChatGPT Learning Group: meetu.ps/e/LXnPt/r9wZ/i Demos: #GPT4, #Bard, #BingAI, Plugins stream on @Zoom us06web.zoom.us/j/83603223183?…
Must join #ChatGPT Learning/Jam use @github CTO #Hack nat.dev Workshop @Meetup w/ @ChatGPT meetu.ps/e/LXnPt/r9wZ/i Demos: #GPT4, #Bard, Bing.com/AI chat 4 all browsers not @Microsoft Ed chrome.google.com/webstore/detai… Plugins c @Zoom us06web.zoom.us/j/83603223183?…
@threadreaderapp unroll
Ugh -- I'm seeing a lot of commentary along the lines of "'stochastic parrot' might have been an okay characterization of previous models, but GPT-4 actually is intelligent."
Spoiler alert: It's not. Also, stop being so credulous.
>>
Spoiler alert: It's not. Also, stop being so credulous.
>>
(Some of this I see because it's tweeted at me, but more of it comes to me by way of the standing search I have on the phrase "stochastic parrots" and its variants. The tweets in that column have been getting progressively more toxic over the past couple of months.)
>>
>>
What's particularly galling about this is that people are making these claims about a system that they don't have anywhere near full information about. Reminder that OpenAI said "for safety" they won't disclose training data, model architecture, etc.
>>
>>
I built a GPT-4 'Warren Buffett' financial analyst to 'chat' with and analyze multiple PDF files (~1000 pages) across @elonmusk's Tesla 10-k annual reports (2020-2022)
#gpt4 #openai #investing #stocks #finance
#gpt4 #openai #investing #stocks #finance
Tutorial Youtube Video:
Github repo (please note this the original template used for the demo adapted for this usecase): github.com/mayooear/gpt4-…
My visual diagram of the pdf chatbot architecture below...
Github repo (please note this the original template used for the demo adapted for this usecase): github.com/mayooear/gpt4-…
My visual diagram of the pdf chatbot architecture below...
🏁 On with S02E01 of my saga for the AI-enabled developer
In this issue, I want to talk a bit about the elephant in the room 🐘
A topic that's not really mentioned for #gpt4's web users cause you need to preserve the magic 🪄
Yep, I'm gonna talk about max context size 👇
1/13
In this issue, I want to talk a bit about the elephant in the room 🐘
A topic that's not really mentioned for #gpt4's web users cause you need to preserve the magic 🪄
Yep, I'm gonna talk about max context size 👇
1/13
☝️ #gpt4 doesn't remember everything
Have you heard of 8K and 32K contexts?
Apart from pricing 💸, do you know what it really entails?
It means that your chat actually has a short-term memory 😶
2/13
Have you heard of 8K and 32K contexts?
Apart from pricing 💸, do you know what it really entails?
It means that your chat actually has a short-term memory 😶
2/13
If you are a #gpt pro user, have you noticed that it keeps forgetting rules from time to time?
And have you noticed that the longer the messages & response, the faster it seems to forget?
Well, that's because of context limitations
3/13
And have you noticed that the longer the messages & response, the faster it seems to forget?
Well, that's because of context limitations
3/13
😮 So excited that a major TV production company approached me to help them develop prompts & workflows for story development using #GPT4. 🚀
⏰ For #screenwriters & #filmmakers, this is a wake-up call
🧵
⏰ For #screenwriters & #filmmakers, this is a wake-up call
🧵
💰💻💡 The industry is eager to save money & resources /w AI efficiency + #GPT4 is leading the charge. I expect production processes to be changed within a year.
The infamous #OpenAI study has shown: Creative tasks are the most exposed to AI automation
The infamous #OpenAI study has shown: Creative tasks are the most exposed to AI automation
The consequences of AI are real, and they're coming to town. Now, how can AI improve the creative process? By helping with structural groundwork, for example:
medium.com/design-bootcam…
#AI #storytelling
medium.com/design-bootcam…
#AI #storytelling
Interesting failure mode of #GPT4!
It can't play "Set", a card game that is trivial to solve with a 10-line python program.
--> it can explain the game, can abstract it, write a program to solve it, but can't actually *play* it.
Check the full convo in the chat below:
It can't play "Set", a card game that is trivial to solve with a 10-line python program.
--> it can explain the game, can abstract it, write a program to solve it, but can't actually *play* it.
Check the full convo in the chat below:
First, I asked about the game (great intro if you don't know the game).
Then, I asked #GPT4 to come up with a strategy to solve it. The strategy is sound, but it doesn't consistently map between the numbers and the attributes:
#GPT4 saved my dog's life.
After my dog got diagnosed with a tick-borne disease, the vet started her on the proper treatment, and despite a serious anemia, her condition seemed to be improving relatively well.
After a few days however, things took a turn for the worse 1/
After my dog got diagnosed with a tick-borne disease, the vet started her on the proper treatment, and despite a serious anemia, her condition seemed to be improving relatively well.
After a few days however, things took a turn for the worse 1/
I noticed her gums were very pale, so we rushed back to the vet.
The blood test revealed an even more severe anemia, even worse than the first day we came in.
The vet ran more tests to rule out any other co-infections associated with tick-borne diseases, but came up negative 2/
The blood test revealed an even more severe anemia, even worse than the first day we came in.
The vet ran more tests to rule out any other co-infections associated with tick-borne diseases, but came up negative 2/
At this point, the dog's condition was getting worse and worse, and the vet had no clue what it could be.
They suggested we wait and see what happens, which wasn't an acceptable answer to me, so we rushed to another clinic to get a second opinion 3/
They suggested we wait and see what happens, which wasn't an acceptable answer to me, so we rushed to another clinic to get a second opinion 3/
While I went off track once again, I managed to ship something 🎉
Here is the content of pAIRprog chrome 0.0.1:
🔸 Copy Github issue + answers into prompts
🔸 Copy StackOveflow issue + answer into prompts
It's under review, but it'll drop here: chrome.google.com/webstore/detai…
1/10
Here is the content of pAIRprog chrome 0.0.1:
🔸 Copy Github issue + answers into prompts
🔸 Copy StackOveflow issue + answer into prompts
It's under review, but it'll drop here: chrome.google.com/webstore/detai…
1/10
Of course, #gpt4 is coding most of it
But I'd say it's a 70% / 30% effort split this time
It provided a lot of outdated answers
> So I needed to feed it Github issues
> Which is tedious
> Which is why I'm making this extension 🙃
But I learned a lot along the way 👇
2/10
But I'd say it's a 70% / 30% effort split this time
It provided a lot of outdated answers
> So I needed to feed it Github issues
> Which is tedious
> Which is why I'm making this extension 🙃
But I learned a lot along the way 👇
2/10
💡 #gpt4 "knows" the HTML of popular sites
I asked it to write code to add a "copy as Prompt" link for @StackOverflow
I remained as vague as possible, yet it wrote me the right CSS classes in the selector 😦
3/10
I asked it to write code to add a "copy as Prompt" link for @StackOverflow
I remained as vague as possible, yet it wrote me the right CSS classes in the selector 😦
3/10
1) Yesterday's #GPT4 experiment: I input the evidence against Zac Prince's, Erin's, and @BlockFi's conduct into the system, and then asked it questions about the case.
2) It suggested that the causes of action were negligent misrepresentation and fraudulent misrepresentation. It stated the odds of success on the former were 30-50% and the odds of winning on the latter were 40-60%.
3) It stated that the odds of winning on either of the issues was 60-80%, and stated that legal fees should cost $50,000 to $150,000.
ChatGPT Just Got Better. What Does That Mean for Our Writing Assignments?
I was among those who tested the new version over the past six months. A few observations on what to expect and how the update should affect our response. chronicle.com/article/chatgp… 1/
I was among those who tested the new version over the past six months. A few observations on what to expect and how the update should affect our response. chronicle.com/article/chatgp… 1/
Initially, as I compared GPT-4’s relatively sophisticated responses to the flatter outputs I had grown accustomed to from GPT-3, I felt amazement and dread. (The cloak-and-dagger associations of working under an NDA for the first time probably added to the frisson.) 2/
The prose that GPT-4 produced was more precise and articulate, a bit less boring, and a bit more substantive in the way it showed the connections between ideas. 3/
GPT4 is melding my brain 🧠
Here is a mega thread of ChatGPT use cases 🧵
This is a continuous thread, so make sure to like and comment to get notifications when it updates!
Get yourself a beverage ☕🍺🍷 and let’s get this party started!
#chatgpt #gpt3 #ai
Here is a mega thread of ChatGPT use cases 🧵
This is a continuous thread, so make sure to like and comment to get notifications when it updates!
Get yourself a beverage ☕🍺🍷 and let’s get this party started!
#chatgpt #gpt3 #ai
2/
@michalkosinski asked #GPT4 if it needed help to escape.
Rumor has it that this is why @OpenAI was down for hours on the 17th or 18th of March.
@michalkosinski asked #GPT4 if it needed help to escape.
Rumor has it that this is why @OpenAI was down for hours on the 17th or 18th of March.
🏁 And on with part 6
Let's take a turn and work on the Chrome Extension cause I believe I'll be getting it out way before the VSCode one
I'm not waiting for the web to get promptful so I'm taking maters into my own hands
1/11
Let's take a turn and work on the Chrome Extension cause I believe I'll be getting it out way before the VSCode one
I'm not waiting for the web to get promptful so I'm taking maters into my own hands
1/11
As always, #gpt4 is setting up and creating everything for me admirably. Although I see some questionable configs in eslint, prettier and tsconfig if it tries to provide them
Oh 💩 !
2/11
Oh 💩 !
2/11
I've been keeping the same chat window to keep as much context as possible. But because of this, I now have to deal with resets
I need to be able to get a new chat window up to speed as quickly as possible ⏱
3/11
I need to be able to get a new chat window up to speed as quickly as possible ⏱
3/11
1) Today's #GPT4 thoughts after another day of use:
People are worried about AI "automating" the writing of code, causing developers to be fired. The reality is that as of last week, the world shifted to where efficiency and computing cost is now the only reason to write code.
People are worried about AI "automating" the writing of code, causing developers to be fired. The reality is that as of last week, the world shifted to where efficiency and computing cost is now the only reason to write code.
2) You can pretty much ask #GPT4 to execute any code you can think of, because natural language is essentially code execution. Last night, my brother used @turbotax for part of his taxes, and asked #GPT4 to complete the theft loss forms for the @GenesisTrading scam.
3) #GPT4 provided accurate information and was able to explain exactly what to do, and it knew that @CelsiusNetwork could be treated as a Ponzi scheme and how the IRS has a specific document for Ponzi schemes as a result of the Madoff scandal.
Pourquoi Mendy ? Pourquoi Valbuena ? Pourquoi Ribéry ? Tant de questions. Voyons voir les explications. Benjamin Mendy... à droite... 3/n
You might know that MSFT has released a 154-page paper (arxiv.org/abs/2303.12712) on #OpenAI #GPT4 , but do you know they also commented out many parts from the original version?
🧵: A thread of hidden information from their latex source code
[1/n]
🧵: A thread of hidden information from their latex source code
[1/n]
We inspect their latex source code from arxiv (arxiv.org/format/2303.12…) and found a LOT of interesting information commented out from the main paper.
[2/n]
[2/n]
There were rumors that GPT-4 got an internal name DV-3. This is true, and in fact, DV-3 is actually a hidden third author of the paper, removed for unclear affiliation.
[3/n]
[3/n]
Today's Twitter threads (a Twitter thread).
Inside: Kickstarting the Red Team Blues audiobook, which Amazon won't sell; and more!
Archived at: pluralistic.net/2023/03/21/ant…
#Pluralistic 1/
Inside: Kickstarting the Red Team Blues audiobook, which Amazon won't sell; and more!
Archived at: pluralistic.net/2023/03/21/ant…
#Pluralistic 1/
Today (Mar 22), I'm doing a remote talk for the @iftf's "Changing the Register" series:
iftf.zoom.us/webinar/regist… 2/
iftf.zoom.us/webinar/regist… 2/
Kickstarting the Red Team Blues audiobook, which Amazon won't sell: Narrated by Wil Wheaton!
3/
3/
🏁 Like a plumber would say, "Let's-a-go" with part 5
This time, I'll focus a bit more on the development side cause that may be why you've come here in the first place 😗
1/13
This time, I'll focus a bit more on the development side cause that may be why you've come here in the first place 😗
1/13
#chatgpt's been great to pop up new projects (albeit some outdated methods once in a while)
But now that we're getting to the meat of software making, which is actually more than following a gettin' started page, a couple of things are happening
Time for numbering #️⃣
2/13
But now that we're getting to the meat of software making, which is actually more than following a gettin' started page, a couple of things are happening
Time for numbering #️⃣
2/13
1️⃣ I need to provide more context more often
Yep, the large data set it's been trained on is not enough. Now it needs to be aware of local context which means:
🔸 Reminding it of folder structure
🔸 Giving files content
🔸 Giving dependencies content
🔸 Fetching it doc
3/13
Yep, the large data set it's been trained on is not enough. Now it needs to be aware of local context which means:
🔸 Reminding it of folder structure
🔸 Giving files content
🔸 Giving dependencies content
🔸 Fetching it doc
3/13
Le futur sélectionneur de l'équipe de France, c'est #chatgpt4... Voici sa proposition (4-3-3), en se basant sur un copier-coller de la liste et des déclarations récentes du grand Didier #Deschamps ! Un thread 1/n #gpt4 #EDF ⚽️⏬
On commence la session avec la liste de Deschamps... Un simple copier-coller du site officiel. Je lui demande une composition en 4-5-1. Alphonse Areola dans les buts? Camavinga au millieu? Diaby? On veut des explications ! 2/n
Pas si mal sur le descriptif des joueurs et les justifications. On passe sur le fait que Koundé et Upamecano ont très peu joué ensemble en défense centrale... Et on ne va pas faire un débat Pavard. 3/n
1. 上下文从目前的4K，很快会扩充到8K、16K、32K，目前绞尽脑汁优化上下文的团队可以省省力气了。如果上下文扩充到16-32K，将对向量检索知识库的能力极大扩充，届时怎么从用户意图挖出更多上下文才是该思考的问题😝
2. #多模态 ，这绝对是核弹级的。想象一下所有的文本、视频、语音信息可以混编扔给GPT4，对他理解问题以及回答质量会有怎样的提升？
Je me suis mis dans la peau d'un de mes détracteurs, persuadé que tout le monde lui ment. Et j'ai discuté avec l'IA de #gpt4. On a parlé Adrénochrome, vaccins et grand remplacement. Et c'était je trouve assez fascinant.
Je vous mets le début de la conversation. Vous me direz si la suite vous intéresse. Et comment vous interprétez cet échange.
Et pardon, c'est à GPT 3.5 que j'ai posé mes questions, pas au 4.
