Pinned Tweet
Follow
Petko D. Petkov
@pdp
security and hacking, sometimes CISO, maker of:
chatbotkit.com 
secapps.com
cohesion.sh
bountypage.com
websecurify.com
gnucitizen.org
secapps.com
cohesion.sh
bountypage.com
websecurify.com
gnucitizen.orgPetko D. Petkov’s Tweets
How many of you are using ChatGPT or equivalent of writing your pentest reports? I wish I had this capability between 2006-2012.
When I was younger, I thought all I wanted in life was to hack and make money. But as I've grown older, I've learned hacking is better enjoyed as a recreational activity, and I find fulfilment in other ways to make money.
1
3
Where do evil scientists get their ideas from? 🤔
Quote Tweet
I was expecting this to happen sooner or latter. We do not need more spam on the Internet.
Quote Tweet
OpenAI has released a way to identify AI vs human-written text!
openai.com/blog/new-ai-cl
Show this thread
3
I completely agree! I was glued to my seat the whole time.
Quote Tweet
Episode 3 of Last of Us is unbelievably good story telling. I am in awe of @Nick_Offerman performance. Incredible.
I am not even the most active user on chatbotkit.com any more. There are customers 100x-ing my usage.
1
So it begins
Quote Tweet
Weird fun fact: GPT natively speaks base64.
Hope your prompt injection checking code is looking for base64!
Show this thread
1
Just ran into this video of Ray Bradbury articulating creativity
It matches exactly how I feel about sitting on a chair using a computer, most of my ideas happen away from the computer
1
7
21
Show this thread
Datasets in chatbotkit.com is kind of like a long-term memory. More information can be found here chatbotkit.com/docs/datasets
1
People leaking their OpenAI keys everywhere. Insane.
2
4
I managed to save a few k of tokens per conversation. chatbotkit.com is getting pretty good at finding the right data based on the specific user context and conversation
1
The results are in
Quote Tweet
I am considering launching a ChatGPT alternative that is privacy and security focused.
Show this poll
42 is a pretty decent number so I am keeping it at that
chatbotkit.com is now consistently punching above its weight - pretty amazing outcome so far
1
We’re taking beta signups for my LLM that de-watermarks LLM output.
Quote Tweet
Incredible paper. LLMs-generated text can be detected by embedding signals that are invisible to humans but algorithmically detectable.
It's a watermarking framework for your language models.
Reaches 99% confidence on 23 words.
arxiv.org/abs/2301.10226
@tomgoldsteincs
arxiv.org/abs/2301.10226
@tomgoldsteincsShow this thread
1
2
21
In my opinion AI will make us more human - in other words we will be able to spend more time on personal growth rather than completing mandate tasks to survive
Quote Tweet
One lesson from AI progress is it’s hard to predict the relative difficulties of various skills; one lesson from AI limitations is that humans are much more capable than we give ourselves credit for.
4
ChatGPT has become almost unusable to me in recent times. Good that I have alternatives.
1
1
I am proud of the tweets I send and the ones I choose not to send, the latter being a larger quantity.
1
Sarcasm aside Mars is just a stepping stone.
Show this thread
Big shoutouts to anyone using LLMs not to generate more Internet spam but to push the boundaries of what this technology can do.
Security is easy when is the first thing you do for your product. It only becomes complicated when it is an afterthought.
1
2
Releasing TRL==0.2.0! 🚀
Training language models with RL (e.g RLHF) is now easier than ever:
New accelerate backend (multi-GPU support!), support for decoder and encoder-decoder models, weight sharing, shiny new docs (and logo), reworked examples!
github.com/lvwerra/trl
🧵:
5
49
196
Show this thread
I call this "inbox 2 months". Everything older than 2 months gets deleted.
I am going through a rebranding exercise based on AI tech only
1
Microsoft Excel is still doing the world a huge favour by allowing exploitable features to be used in 2023. We in the InfoSec world can only give Microsoft a thumbs up for making our jobs so interesting and secure.
1
1
I hear you all. I am very excited to share #Algo by ChatBotKit - the ultimate ChatGPT alternative built with privacy and security as top priorities. Experience the power of AI without compromising your data. Try it now and discover the difference. algo.chatbotkit.com
3
Show this thread
Ha, nothing would surprise me less than AWS and Apple coming up with some kind of revolutionary move for Alexa and Siri.
1
Anyone knows of any good research on prompt injection defence techniques?
1
1
Still improving on the prompts and UI but it is fully functional and general enough to be used instead of ChatGPT. Some cool features coming soon such as the ability to access web pages, use custom datasets and more.
Show this thread
🚀 Exciting news! Just launched a ChatGPT alternative with added privacy features 🔒 Check it out now on ProductHunt! producthunt.com/posts/algo-by- 🤖 #ChatGPT #privacy #AI 🚀
2
7
Show this thread
Google has lost the plot. Even their search indexing capabilities are now far outpaced by Bing and Duck.
1
1
5
From a 1979 IBM presentation
100
1,569
7,988
Anyone has any examples of good chat UI?
2