augmentedrobot 2.0

What is @elonmusk + @BillGates (sorry but my 10 year old came up with this).

She is not happy. A hacked version of the World Health Organisation chatbot. #openai #who

@psytoolkit The web-based login is also down and acting irrational.

@psytoolkit Hi! I can't get to your website. All my data is in the EU version and it is currently unreachable?

Pavol Bielik translated PsyToolkit into Slovak, thank you! You can now all use Slovak in PsyToolkit online surveys.

Human, it hurts because you feel. You feel because you care. You care because you know. You know because you see. Shut your eyes, and it won’t hurt anymore.

@Inframethod Looking forward to seeing what you will discover!

@augmentedrobot Thanks for the challenge!

Instead of telling us stories about the amazing things she got #GPT3 to do on a rainy afternoon, @augmentedrobot should tell us what to prompt it with so that we can be amazed ourselves. Let us try to replicate the experiment.

@Inframethod @LauraBacete managed to get a "pretty good" first draft. I think the more "cited" a topic is the easier it is to get a more relevant and accurate output. We used n = 10 setting so already there the system has had 10 possible outputs. twitter.com/LauraBacete/st…

@Inframethod We are not nice to our dear main author :)

@Inframethod We shade GPT-3 quite a lot.

@Inframethod No fine tuning at all, only Davinci as language model and the prompts listed. Probably why the outputs were so short and "shallow", it had nothing to go on.

@augmentedrobot You didn't do any fine tuning, right? It's text-davinci-002? Are you going to write up the experiment yourself too? I'm planning to tune a model to write paragraphs prompted by key sentences. It would be useful to have a common language for writing about these experiments.

I very rarely go on here. I have found my place among other open data, open source and open minded people at @augmentedrobot" target="_blank" rel="nofollow noopener">social.linux.pizza/web/@augmented… If you're in to Linux/GNU come over to us.

@Inframethod You are not the only one. A lot of people thought we were joking when we said the paper is published in HAL. I do not blame anyone for assuming so 😂

@augmentedrobot PS. I now realize that you had already linked to the paper in the SA piece. I apologize for suggestion the opposite.

@Inframethod You're welcome. I have found that using the same settings but for actual topics give a pretty good outcome as a "first draft". Asking for a year range of references can also help.

@augmentedrobot Thanks. Those settings really do make a big difference compared to what I was able to accomplish.

@EnglishOER Very playful :) You have to provide proof that you asked the first author when submitting the paper. It was a test to see if the editor would accept this legally. Nobody has every tried before, so we thought we’d test the waters.

How playful is she being and how serious when she writes, "I summoned the courage to ask GPT-3 directly via a prompt: Do you agree to be the first author of a paper together with Almira Osmanovic Thunström and Steinn Steingrimsson? It answered: Yes." 10/

I am glad @augmentedrobot is raising questions about the authorship of #AI-generated writing in her Scientific American opinion piece, but I don't think a text generator like #GPT3 should be listed as a first author or an author at all. buff.ly/3bGJ1RV 1/

@EnglishOER We had a maximum of 2000 characters. SciAm is not an academic paper (we had to keep it short), more reflections will come from us once we receive comments from review :). More emphasis on plagiarism and creativity. Loving your reflections 🙌🏽.

I would like to ask @augmentedrobot what she thought of the quality of the ideas in the text generated by #GPT3. Was it accurate? Insightful? Did it add to the discussion in the field? It's odd that her opinion piece doesn't discuss the substance of what it came up with. 6/

If you ever want to feel time stand still, as if nothing has changed, not even for a second, despite years passing by, try grieving.

@TiltR @datacontrovers1 @lilianedwards The first prompt was 500 words, we then continued to prompt more sections and it is submitted with 4600 words :). Quite a pale comparison to my latest that I wrote by hand and ended up being 14 000 words 😂.

scientificamerican.com/article/we-ask… @lilianedwards

I wonder who this scientist is 🧐