Less protocol-complaint but likely also dumber. You see a similar phenomenon if you ask it to speak in l33t5p34k, or with hyphens between all letters, or especially with Zalgo text. Base64 is interesting though because it probably sees a *lot* of translation pairs in pre-train.
AI
-
text-davinci-002 lacks RLHF; ChatGPT belongs to 002 line
By
–
We now know that text-davinci-002 doesn’t use RLHF either, and based on the model names used in API calls ChatGPT appears to be part of the 002 line, rather than 003 which uses RLHF.
-
Prompt injection shatters ChatGPT’s coherent personality illusion
By
–
You can prompt-inject ChatGPT into behaving however you like, but this shatters the illusion of a coherent personality. ChatGPT forces you to type your own disclaimer to prove you understand it’s not real. Otherwise you talk to the protocol droid.
-
DeepMind UCL Fine-tune 70B LM for Human-aligned Statements
By
–
DeepMind & UCL Fine-tune a 70B Parameter LM to Generate Statements Agreeable to Humans with Diverse Opinions
-
AI Enables Rapid Code, Content Creation, and Idea Testing
By
–
Generating code, building a website, automating tasks, and speeding up the production and creation of content were either expensive or time-consuming but necessary hurdles that can now be generated in seconds The cost and speed to test a business idea will approach zero with AI.
-
AGI Ticket NFT: Free Collectible Rewards Historical Debate Participation
By
–
"AGI TICKET NFT" FREE 0 Ξ (+ Transaction Fee) A Collectible #NFT to reward historical #AGIDebate participation. MINT WEBSITE: https://
app.manifold.xyz/c/agiticket #AGI #AGIDebate #AIDebate #AGITicket #NFT -
HyperWrite AI Tool Improves Style Emulation Capabilities
By
–
Yes! As you use HyperWrite, it’ll continually get better and better at emulating your style. Next year, we plan to do a deeper integration like you mentioned.
-
Fast and Slow Thinking for Autonomous Robot Navigation
By
–
Combining Fast and Slow Thinking for Human-like and Efficient Navigation in Constrained Environments Ganapini et al.: https://
arxiv.org/abs/2201.07050 #ArtificialIntelligence #DeepLearning #MachineLearning -
ChatGPT correctly answers Super Bowl question about Justin Bieber’s birth year
By
–
Here's a related tweet, from the pre-ChatGPT era. Nowadays, ChatGPT can correctly answer the question, "What NFL team won the Super Bowl in the year Justin Bieber was born?" without any CoT prompting, unlike prior GPT-3 versions.
-
Correct placement of ‘Let’s think step by step’ in prompts
By
–
It's also widely misunderstood as an instruction suffixed to the question, vs. a prefix to the answer. The correct format is: Q: A: Let's think step by step. I suspect it works specifically because it precludes the answer form you described.