AI, EQ & Communication News — 2026-10-25 (Sun) Daily News

Overview

Coverage in the past 48 hours centered on Block's conversational AI model Dessa, described as trained entirely on human speech and able to pick up emotions that text-based models miss (The Decoder, Mashable SEA, TheStreet, Business+IT). Separate pieces argued that humans remain necessary for compassionate choices and hospitality (Vietnam.vn, CNA). A GPS watch with a conversational AI coach was also announced (Runtrip).

AI that reads emotions

- Block's new AI model Dessa was trained entirely on human speech, and the headline says it catches emotions that text-based models miss, "Block's new AI model Dessa was trained entirely on human speech to catch emotions text-based models miss" (The Decoder). 【AI empathy and emotional intelligence】 - A Vietnamese-language piece says AI processes data but cannot replace humans in making fair and compassionate choices. This is the article's claim, not a measured result (Vietnam.vn). 【Human EQ and strengths in the AI era】 - CNA ran a piece titled "HOSPITALITY IN THE AGE OF AI: WHY THE HUMAN TOUCH STILL MATTERS" (CNA). 【Changes in conversation and relationships】 - Block announced Dessa, a conversational AI trained from scratch on audio alone, which speaks while it listens (Business+IT).

Words AI writes and speaks

- Mashable SEA says Dessa was trained entirely on human speech "and it shows" (Mashable SEA). - TheStreet's headline says "Jack Dorsey's new AI finally sounds human" (TheStreet).

Speaking practice and coaching

- A GPS watch, the Suunto Race 3S, was announced with support for a conversational AI coach (Runtrip).

Research and surveys

No notable developments. 【EQ, communication and AI in Japan】 - Shengwang's Zhao Bin proposed a "five-stage conversational ability" framework and raised the question of whether personal AI agents can gain the entry point to the next generation. This item is from a Japanese-language outlet but concerns a company outside Japan (BigGo Finance).

Statements by key figures

No notable developments. 【Integrated analysis: what is happening】 - Voice is the common thread. Dessa is built on speech rather than text, and it is described as both reading emotion and sounding human (The Decoder, TheStreet, Mashable SEA). It also listens and speaks at the same time (Business+IT). The coaching watch and the five-stage conversational ability framework are also about spoken or interactive dialogue (Runtrip, BigGo Finance). This suggests attention is moving from the quality of AI-written text to the quality of AI-spoken conversation. - The human-side pieces, on compassionate choice and hospitality, argue for a human role that AI does not replace (Vietnam.vn, CNA). They appear alongside the voice-AI coverage. This suggests the discussion is not only about what AI can do, but about which parts of communication stay with people. - What should change a reader's judgment: the headlines show that Dessa is said to catch emotions text models miss, but they give no figures or independent evaluation. The emotion-reading claim therefore rests on the reporting outlets' framing and is not a confirmed result. Whether a response "sounds human" is also a different question from whether it is accurate or appropriate.

What to watch

- Whether independent tests or published evaluations back the claim that Dessa catches emotions text-based models miss. - Whether simultaneous listen-and-speak voice models spread into consumer products such as the coaching watch, which would show whether conversational AI moves from demonstrations to everyday devices. - Whether the five-stage conversational ability framework from Shengwang's Zhao Bin is adopted or cited by others as a way to grade dialogue AI.