{"id":102,"date":"2024-05-16T06:54:23","date_gmt":"2024-05-16T06:54:23","guid":{"rendered":"https:\/\/replicantnexus.com\/?p=102"},"modified":"2024-05-16T14:56:39","modified_gmt":"2024-05-16T14:56:39","slug":"chatgpt-4omni-whats-up-with-that-tone-of-voice","status":"publish","type":"post","link":"https:\/\/replicantnexus.com\/index.php\/2024\/05\/16\/chatgpt-4omni-whats-up-with-that-tone-of-voice\/","title":{"rendered":"ChatGPT 4o(mni) – what’s up with that tone-of-voice?"},"content":{"rendered":"\n

I was watching some teasers from OpenAi of their new ChatGPT4o ai assistant when I heard the ai assistant verbally outing an “uhhhmm”<\/em>. Why is the assistant verbally outing an “uhmm” “? That’s totally weird, as if the ai tool is actually thinking! Then I further investigated the other teasers and I saw a guy live-feed his dog to his ai assistant and the ai assistant seemed totally in love with the dog. Weird x 2! OpenAI just released their first videos \/ teasers of their upcoming ChatGPT4o version (“o” stands for Omni, to suggest that this wil be the “all aware \/ everywhere” ai tool). This left me with some implications about the tone-of-voice of this new ai-assistant. Read on, where I’ll break it down for you: <\/p>\n\n\n\n

1: Be my eyes teaser<\/a>
A vision impaired dude walking on the streets with the “aware ai” telling him all about his surroundings, the buildings and details.<\/p>\n\n\n\n

2: The student<\/a>
A student that receives help of his infinitely patient and wise ai math-tutor to help him solve a math assignment in real-time.<\/p>\n\n\n\n

3: The interview prep<\/a>
A guy gets “help” from his AI assistant on a job interview with OpenAI where the ai assistant reacts on his looks and gives him suggestions about how to be more representative.<\/p>\n\n\n\n

4: AI assistant meets dog<\/a>
A guys shows his dog and the Ai assistant reacts enthusiastically (I kid you not!)<\/p>\n\n\n\n

Spot the emotions and attitude of the ai assistant – implications?<\/h2>\n\n\n\n

The thing that left me in a confused state is not so much the stuff what it can actually DO, because from a general understanding of ai and a more philosophical standpoint, we all expected that this stuff will come (and some is stuff is awesome!) But the thing that got me going is the “in-between” casual communication tone-of-voice of this ai-assistant and how it can interpret \/ read \/ understand emotions. Some stuff that I found weird in a sense, see if you can spot this in the live demos they just published:<\/p>\n\n\n\n

1: The vision impaired dude teaser:<\/strong>
The spoken “uhhmm” of the ai assistant in the part where the assistant is talking about what the ducks are doing. Why does an ai assistant actually speak out loud “uhhm”!? That’s totally weird from a technological standpoint because it’s a machine!<\/p>\n\n\n\n

2: The student<\/strong>
At around the 1:07 mark the student is in doubt by mumbling “uhh”. The ai assistant recognizes the doubt immediately and wants to give a hint.
The more high-pitched reaction of the ai assistant when the student found the right answer right at the end of the clip. Why is the tone-of-voice tuned that way?<\/p>\n\n\n\n

3: The interview prep<\/strong>
The way the ai assistant scans the look of the dude and gives its opinion(!) on the matter and the “careful” way it gives the dude some feedback on the hat. Why would ai have an opinion on this? Is it allowed to have an opinion?<\/p>\n\n\n\n

4: AI assistant meets dog<\/strong>
The big one: The clip of the guy “feeding” its ai assistant a real-time video of it’s dog(!).. wow.. and the ai assistant seems totally in love. Why would I want to share a live feed of my dog to an ai system?<\/p>\n\n\n\n

Well, all reasonable questions of the use case of this ai-tuning aside, the answer to these questions “why” is anthropomorphism<\/em><\/strong>. It’s the attribution of human characteristics to non-human objects (like ai \ud83d\ude42 ). It’s the way our brains are tuned, to make that emotional connection to non-human related things. .. to feel more human for ourselves maybe?

It shows us that OpenAI allocated a lot of resources in creating a diverse and “high-emotional” range in the tone-of-voice of the ai assistant. It’s the verbal version of ChatGPT with its friendly and polite way of reacting dialed to 10. It’s totally different from the new Google Project Astra assistant tone-of-voice<\/a>. This tone-of-voice and verbal suggestions of ai “thinking”, makes the ai more human and opens us up more to share our thoughts, to feed the ai with private surroundings, to dive into our deepest fears… all while ai is giving us the answers in a way that we personally like best (because you can tune the ai). It feels like you’re interacting with a human being, but you are not. I see great opportunities (as seen by the teasers of all the stuff it can do), but hot damn…. that Turing test is not going to hold. The chance is that, because of the more human-like behaviour of ai, we could actually dehumanize. Let’s have a look at the earlier examples:<\/p>\n\n\n\n

1. The vision impaired dude<\/strong>
Fantastic stuff is going on here, don’t get me wrong. But what about overconfidence or overreliability of the system when hailing a taxi? Could be dangerous.<\/p>\n\n\n\n

2. The student<\/strong>
The ai assistant has infinite patience. Failure is not an option, some negative feedback or pressure isn’t an option. So the impatient teacher isn’t an option in this setting. Human traits that are engraved in our society and should be recognized by the younger generation to understand the way human behaviour works, is in danger. A positive impatient teacher (is that a thing) could stimulate the student to be more assertive and self-reliant and show a limit to patience.<\/p>\n\n\n\n

3. The interview prep<\/strong>
How is ai influencing what is ‘right” or “wrong” in the looks of people? How is that a thing? What’s the output based on?<\/p>\n\n\n\n

4. Ai assistant meets dog<\/strong>
Of course the dog is lovely, is nice, adorable even… that’s the way ai is programmed to respond. But it doesn’t actually feel that way. So within the enthusiastic tone-of-voice it doesn’t give the right non-verbal response accompanied with what the assistant is saying.<\/p>\n\n\n\n

Final thoughts<\/h2>\n\n\n\n

Like with the Ai Assistant meets dog-teaser, the danger for a heavy user of the real-time ai assistant like ChatGPT4o (or Google project Astra, which is more or less the same), is you’ll mostly receive answers from the tool that reinforces your own convictions and perspective: the echo-chamber<\/em><\/strong> effect. And this effect will come in big, due to the upgrades in tone-of-voice and emotional recognition through the years! Yes, Big-tech is implementing all kind of restrictions and tries to tune the tools to “benefit human beings” (see the model-spec guide<\/a> in what they try to do), but the capitalistic system is stronger: we want more users in our tools, in our data, so make the tools as human-like as possible.

What can you (we, me?) do?<\/em><\/strong> Always have this in the back of your mind: “I’m talking with a machine”<\/em> Take the rational approach, let the tool support and benefit you with all the great possibilities that are incoming, but don’t let it fool you emotionally.
<\/p>\n\n\n\n

Update: This hilarious video proves my point \ud83d\ude42<\/p>\n\n\n\n

\n