AI detector credits Bible to ChatGPT 🤦

🎓 Featured Story:

Students around the world are now using ChatGPT to complete their assessments. Education institutions around the world are trying to deal with this new kind of 'cheating' using AI content detection software.

The problem?

• Differentiating between AI and human created work is hard for plagiarism detection softwares

• Bible’s initial paragraphs are incorrectly flagged as AI-written 🤦 when I pasted them to ZeroGPT, one of the most popular AI content detectors.

• Numerous students are being wrongly accused of plagiarism because of these unreliable detectors.

🧠 Key Insight:

Major redesign of assessments and methods of submission would be needed by education institutions in the long-term to combat this problem. It would not be sustainable to fight off technologies like ChatGPT or to keep using these unreliable tools. [link]

Other updates to keep you ahead of the curve:

1. 📷 Zoom now offers AI summaries for missed meetings. [link] They are available in select plans as free trials for now.

2. 🏛️ OpenAI sued after ChatGPT wrongly accused a radio host for embezzling funds. [link] Insight: Do you think it is the user’s responsibility to know that ChatGPT can just make up information or is it OpenAI’s responsibility to make sure it doesn’t spread false information?

3. ✍️ ChatGPT is making copywriters unemployed. Companies know that they’ll get worse results, but consider them worth the price drop anyways. [link]Insight: Tech is no longer automating only repetitive jobs but even high paid or white collar ones.

4. 🔍 Google has added new math, coding and logic abilities for its chatbot Bard. [link] This is important because traditionally chatbots have been bad at this stuff and only been good at language.

5. 🎻 Meta released its open-source music generation model. You can give it a prompt or audio sample and it will produce relevant 12 seconds of audio. It’s good but won’t be automating musicians any time soon. [link]

6. 🎨 AI now allows you to generate incredible artsy QR codes. [link] Yes, this art is a fully functioning QR code that you can scan with your phone!

⚒️ AI Toolbox: ChatGPT & Prompt engineering (Part 2)

To recap,

• Sometimes ChatGPT won’t work if you don’t ask your question the way “it likes”.

• Prompt engineering is this skill of phrasing the question well to language models like ChatGPT get best answers. [Part 1]

Here are some tricks to ask questions to ChatGPT -

→ Ask ChatGPT to solve a problem by providing reasoning or steps

For anything complex, ask ChatGPT to work out the problem step-by-step or provide reasons rather than asking to answer concisely.

❌ Initial Prompt: Asking to answer in one word gives INCORRECT answer.

✅ Better Prompt with step by step thinking: Asking to answer with steps gives CORRECT answer.

→ Ask ChatGPT “Are you sure?” after it replies

This is a technique I use on a daily basis when using ChatGPT for programming. When I suspect it has suggested something wrong, I ask “are you sure?”. It then usually corrects itself.

→ Ask ChatGPT things it must follow rather than what it must avoid

ChatGPT understands well when you tell it to do something specific rather than when you tell it to NOT do something [link]

❌ Initial Prompt with what NOT to do: (It doesn’t actually follow)

✅ Better Prompt with what to do: