TweetMirror - Train a LLM on your Tweets

Hacker News

Hi HN, Tweet Mirror is a toy project I made to learn about Large Language Models. With the speed that the field is advancing, I felt I'd be remiss if I didn't learn more about it. To use it, you log in with Twitter, pay the amount it costs to train the model (GPUs are expensive otherwise I'd release this for free!) and wait 15m to an hour for fine-tuning to receive tweets generated in your style. Technically, I fine tune the GPT-J 6B parameter model for a heuristically determined number of epochs with early stopping. I also perform a lot of data preprocessing/sanitization which improves results. In addition, I perform topic modeling using a pipeline similar to BERT-topic and condition some % of your tweets on common topics as well as what is more liked. With this project, I wanted to engender conversation around a couple of topics that came to mind while building it. Curious for all of your thoughts and feedback at large! - Economics and defensibility of building AI. I wanted to release this for free but had to charge because it costs a non-trivial amount per job. If fine-tuning models / training models for users will be valuable in many applications, how will this shape the business models of companies that use these technologies? Is there something intrinsic about the cost structure that gives benefit to incumbents? - Dangers of "loose" AI agents. There is nothing stopping a bad-actor from doing what I have done here and letting it loose on the internet, perhaps even with RL as feedback. Thus creating a seemingly real person that runs on its own. Countries such as Saudi Arabia have already used twitter bots for their agenda, and this technology seems much more powerful than what they used. Should we regulate AI agents being released "into the wild". How do we navigate AI that has convincing agency on the internet, which feels very feasible in the near-term? - There is something really spooky about the generated tweets because they often share your vernacular and sound quite similar to you. I think this raises interesting philosophical questions. What does it mean to be human in a world where your thought process can be git-forked? Curious what you all think, and how your Tweet Mirror's make you feel!

Visit product View on Hacker NewsNov 29, 2022

Share card

Actual performance

4points

Did not reach leaderboard

Launch Intel predictions

Analyze your own launch →

Product HuntOn track for Day 1 leaderboard · Strong signals: agents, agent, model · Missing: mac, macos, cursor

89%89% predicted probability of success on Product Hunt, based on ML models trained on real launch data.

best fitHighest predicted score across all platforms for this description.

Indie HackersFits the IH revenue-focused audience · Strong signals: para · Missing: supports, reddit linkedin, podcasting

78%78% predicted probability of success on Indie Hackers, based on ML models trained on real launch data.

Hacker NewsMay not resonate with HN audience · Strong signals: lua, pipe, io · Missing: https docs, excited, just released

43%43% predicted probability of success on Hacker News, based on ML models trained on real launch data.

nativeThis product was originally launched on this platform.

AppSumoMay struggle as an AppSumo deal · Strong signals: users · Missing: plus, platform, intuitive

39%39% predicted probability of success on AppSumo, based on ML models trained on real launch data.

TrustMRRLess likely to generate early MRR · Strong signals: users, para · Missing: mobile apps, ios, personal

32%32% predicted probability of success on TrustMRR, based on ML models trained on real launch data.

Acquire.comPre-revenue stage for this audience · Strong signals: training · Missing: arr, mrr, revenue

18%18% predicted probability of success on Acquire.com, based on ML models trained on real launch data.

BetaListMay not resonate with beta-testers · Missing: web3, chat, crypto

0%0% predicted probability of success on BetaList, based on ML models trained on real launch data.

Correct prediction on native model

Similar products

Train an LLM with Metaflow, Featuring Dolly39%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Train an LLM with Metaflow, Featuring Dolly

Hacker News3

Tweets Sentiment Analysis with LLM42%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Tweets Sentiment Analysis with LLM

Hacker News3

Framed Tweets37%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Framed Tweets

Hacker News2

Hurricane Tweets37%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Hurricane Tweets

Hacker News56

Beyond Tweets37%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Beyond Tweets

Hacker News1

Sex Your Tweets (Bayesian classifier)37%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Sex Your Tweets (Bayesian classifier)

Hacker News1

Find Haiku in Your Tweets37%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Find Haiku in Your Tweets

Hacker News1

Lonely Tweets :'(42%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Lonely Tweets :'(

Hacker News1

#e is for Ephemeral – timed deletion of your tweets55%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

#e is for Ephemeral – timed deletion of your tweets

Hacker News1

A game made out of tweets36%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

A game made out of tweets

Hacker News2

Browse all products Analyze your launch