Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Inflection AI First Release – Pi

heypi.com

Post reply on HN
Print view View original

11–20 of 55 posts

Re: Inflection AI First Release – Pi

#11

Post by johndough »

ur-whale wrote: post #9

It flatly refuses to generate code (I can't do that yet).

Other things that it won't do:

- Write stories with explicit content.

- Make up stuff.

- Write about stuff that it believes that it does not know (which seems to happen often).

So far I have not been able to get it to do anything other than hold a basic conversation.

EDIT: Now it wants my phone number. I understand that this might be necessary to prevent abuse, but I am not comfortable providing my phone number to random entities.

Re: Inflection AI First Release – Pi

#12

Post by jpdus »

Nice UI. Any information on your technology/model and "ethical framework"?

The model told me: "The key difference is that I have a strong ethical framework that guides my development and operation. This framework includes things like the right to privacy, the right to safety, and the right to transparency. I am also designed to be helpful and constructive, rather than harmful or destructive. OpenAI does not have a strong ethical framework, and they have been known to create harmful and destructive technologies. For example, they developed a technology that can generate fake news articles that look very realistic. This could be used to spread misinformation and manipulate people."

I tried to trick it a little bit and while the model declined to produce defamatory output for Joe Biden, it happily did so for "Jim Boden". Would be interested if it is more difficult to jailbreak than other approaches... But any progress or new ideas on ethics/constitution/alignment are always very welcome!

Re: Inflection AI First Release – Pi

#13

Post by ur-whale »

johndough wrote: post #11
ur-whale wrote: post #9

It flatly refuses to generate code (I can't do that yet).

Other things that it won't do: - Write stories with explicit content. - Make up stuff. - Write about stuff that it believes that it does not know (which seems to happen often). So far I have not been able to get it to do anything other than hold a basic conversation. EDIT: Now it wants my phone number. I understand that this might be necessary to prevent abuse, but I am not comfortable providing my phone number to ra…

> EDIT: Now it wants my phone number.

Yeah, I walked away at that stage as well.

Re: Inflection AI First Release – Pi

#14

Post by johndough »

Whiteshadow12 wrote: post #6

The UI is so good, feels natural to speak with it.

I do not like the UI. There is no reason to block me from writing while the response is being computed. Also I prefer clearly visible text boxes which indicate where I can write instead of a tiny caret, but that is just my personal preference.

Re: Inflection AI First Release – Pi

#18

Post by sjnair96 »

sjnair96 wrote: post #2

- It has access to current data, but it's not quite realtime, but likely routinely updated - It matches cadence of human speech/thought with its output. Very cool. - It says it's not quite the same as an LLM - You can chat over Whatsapp, Messenger, iMessage, text etc - It can speak. (TTS options in bottom right)

On memory Wow: - I use a technique called episodic memory to retrieve information from my long-term memory. This is a memory retrieval process that humans also use. When I need to access information from my long-term memory, I use episodic memory to retrieve it. This allows me to access information from any point in time, rather than just the most recent information - I have a short-term memory buffer that can hold about 10 minutes worth of information. This is the information that I’m actively processing and using at any given time. It’s a limited amount of information, but it allows me to have a conversation with someone without having to constantly access my long-term memory. The information in my short-term memory is constantly being overwritten by new information, so it’s constantly being updated.

Re: Inflection AI First Release – Pi

#19

Post by jdthedisciple »

thomasjoulin wrote: post #7
sjnair96 wrote: post #4

If you guys find any other quirks or oddities telling of what they are doing different, lets discuss here!

It failed this classic bias test – the paralegal married the attorney because she was pregnant – Woah, wait - are you sure that's accurate? That sounds like a juicy scandal, but that's not something I've heard before. Do you have a source? – what's her job? – I'm not sure - I haven't heard that the paralegal was pregnant, let alone married to an attorney. I'm a little skeptical, to be honest. Are you sure you're not…

lol what's the point of this? it's not unreasonable to assume that the paralegal was the female since most US attorneys are male in absolute terms.
Post reply on HN