正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

We’ve spent years collecting & annotating speech data. What became equally important was learning how to evaluate speech models in a way that reflects real listeners, real accents, & real usage. This video explains how we ran this blind evaluation & what Josh Talks is building in speech evaluations.

Shobhit Banga

2,998 subscribers

24,026 次观看 • 6 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

"In what way is this credible?": Irish Justice Minister Helen McEntee is asked how activists and government-funded NGOs who have campaigned for hate speech laws for years can be cited as evidence that her hate speech bill is popular. #gript

"In what way is this credible?": Irish Justice Minister Helen McEntee is asked how activists and government-funded NGOs who have campaigned for hate speech laws for years can be cited as evidence that her hate speech bill is popular. #gript

gript

1,218,138 次观看 • 3 年前

The Toronto Star instilled hatred and never apologized. Shameful. This is what you call real hate speech from a major Media source in Canada.

The Toronto Star instilled hatred and never apologized. Shameful. This is what you call real hate speech from a major Media source in Canada.

🇨🇦 Antonio Tweets

26,319 次观看 • 11 个月前

Daniel Boguslaw How is this any different from what happens in Iran? Freedom of speech is in danger and so are we.

Daniel Boguslaw How is this any different from what happens in Iran? Freedom of speech is in danger and so are we.

Former Republican 🇺🇸

106,132 次观看 • 1 年前

In this 1997 commencement speech, I offered a warning that America's growing economic divide would undermine the common good. If you're curious how that speech holds up 28 years later (or what I look like without a beard), watch this.

In this 1997 commencement speech, I offered a warning that America's growing economic divide would undermine the common good. If you're curious how that speech holds up 28 years later (or what I look like without a beard), watch this.

Robert Reich

41,821 次观看 • 1 年前

Fadiel Real Deal Adams first Speech as an MP in 2024 was 🔥🔥🔥 I wonder what happened to this brother few months after this speech His party was going to be bigger than PA... But he got entangled in the SAPS CI issues

Fadiel Real Deal Adams first Speech as an MP in 2024 was 🔥🔥🔥 I wonder what happened to this brother few months after this speech His party was going to be bigger than PA... But he got entangled in the SAPS CI issues

Schwarzenegger Cawuke

22,105 次观看 • 1 个月前

bryan’s speech to amaya was exactly what she needed to hear after everything she went through in that villa. she really found a real man #loveislandusa

bryan’s speech to amaya was exactly what she needed to hear after everything she went through in that villa. she really found a real man #loveislandusa

dai

977,575 次观看 • 1 年前

I just spent over 2 hours in Connie Bapela Chambers listening to what sounded more like a political campaign speech than a Budget Speech. Walking out, I noticed a broken window inside the chambers. A building not even 10 years old already reflects the decay of Joburg itself.

I just spent over 2 hours in Connie Bapela Chambers listening to what sounded more like a political campaign speech than a Budget Speech. Walking out, I noticed a broken window inside the chambers. A building not even 10 years old already reflects the decay of Joburg itself.

Herman Mashaba

30,587 次观看 • 3 个月前

#EditorsPeek: When the word ‘Cogress’ was used 50+ times, was the PM’s speech yesterday an ‘official address to the nation’ or a political speech? Watch this video as Priyanshi Sharma explains how such a speech has almost never been made in an 'official address' by any Indian PM.

#EditorsPeek: When the word ‘Cogress’ was used 50+ times, was the PM’s speech yesterday an ‘official address to the nation’ or a political speech? Watch this video as Priyanshi Sharma explains how such a speech has almost never been made in an 'official address' by any Indian PM.

Peek TV

20,985 次观看 • 4 个月前

Welcome to Hollywood’s War on Free Speech Free Palestine or Pro israel... In Hollywood, pick a side and get canceled, shamed, or boycotted. But here's the thing - silencing people kills free speech and shuts down real dialogue. Is that really what we want?! Share this video if you agree!! Follow buildersofmideast for more!🙌

Welcome to Hollywood’s War on Free Speech Free Palestine or Pro israel... In Hollywood, pick a side and get canceled, shamed, or boycotted. But here's the thing - silencing people kills free speech and shuts down real dialogue. Is that really what we want?! Share this video if you agree!! Follow buildersofmideast for more!🙌

buildersofmideast

13,589 次观看 • 1 年前

Watch "Endless experimentation: Building AI models in the wild." This is a 50-minute, MIT lecture about problems when deploying models and LLMs in the real world and how to prepare to solve them. This is a great lecture for those building production machine learning.

Watch "Endless experimentation: Building AI models in the wild." This is a 50-minute, MIT lecture about problems when deploying models and LLMs in the real world and how to prepare to solve them. This is a great lecture for those building production machine learning.

Santiago

117,204 次观看 • 2 年前

I taught a speech model to understand context in conversation. This is what happened It adjusts voice and tone to express urgency, comfort, understanding from the dialogue. Just like a real human being 520M model. Runs locally on consumer devices How this is achieved 🧵

I taught a speech model to understand context in conversation. This is what happened It adjusts voice and tone to express urgency, comfort, understanding from the dialogue. Just like a real human being 520M model. Runs locally on consumer devices How this is achieved 🧵

Luozhu

21,944 次观看 • 6 个月前

We’re excited to introduce KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI, accepted at #ICASSP2026! 🐢 Blog Paper Can a speech AI think deeply without pausing to process? In real conversation, we don’t wait until we’ve fully worked out what we want to say—we start talking, and our thoughts catch up as the sentence unfolds. Fast speech-to-speech models achieve this, but their reasoning tends to stay shallow. Cascaded pipelines that route through a knowledgeable LLM are smarter, but the added latency breaks the flow—they fall back to "think, then speak." In our new paper, we propose a way to break this trade-off. We call it KAME (Turtle in Japanese). A speech-to-speech model handles the fast response loop and starts replying immediately. In parallel, a backend LLM runs asynchronously, generating response candidates that are continuously injected as "oracle" signals in real time. This shifts the AI paradigm from "think, then speak" to "speak while thinking." The backend LLM is completely swappable. You can plug in GPT-4.1, Claude Opus, or Gemini 2.5 Flash depending on the task without changing the frontend. In our experiments, Claude tended to score higher on reasoning, while GPT did better on humanities questions. Try the model yourself here:

We’re excited to introduce KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI, accepted at #ICASSP2026! 🐢 Blog Paper Can a speech AI think deeply without pausing to process? In real conversation, we don’t wait until we’ve fully worked out what we want to say—we start talking, and our thoughts catch up as the sentence unfolds. Fast speech-to-speech models achieve this, but their reasoning tends to stay shallow. Cascaded pipelines that route through a knowledgeable LLM are smarter, but the added latency breaks the flow—they fall back to "think, then speak." In our new paper, we propose a way to break this trade-off. We call it KAME (Turtle in Japanese). A speech-to-speech model handles the fast response loop and starts replying immediately. In parallel, a backend LLM runs asynchronously, generating response candidates that are continuously injected as "oracle" signals in real time. This shifts the AI paradigm from "think, then speak" to "speak while thinking." The backend LLM is completely swappable. You can plug in GPT-4.1, Claude Opus, or Gemini 2.5 Flash depending on the task without changing the frontend. In our experiments, Claude tended to score higher on reasoning, while GPT did better on humanities questions. Try the model yourself here:

Sakana AI

293,975 次观看 • 4 个月前

Where does free speech end—and real danger begin? In this clip, I explain how the legal standards around speech have evolved in America, and why today’s environment makes it harder than ever to define the line between protected expression and incitement. I walk through how the First Amendment has been interpreted over time—especially with changes introduced by the courts that expanded protections for speech, even when it’s offensive or extreme. But here’s the real issue: not all speech is equal. We already limit certain types—whether it’s commercial speech, explicit content, or harmful material—and yet when it comes to rhetoric that can inflame or incite real-world harm, the standards become murky. I raise the question: what happens when speech crosses into something more dangerous—and who decides? This isn’t about silencing debate—it’s about understanding responsibility in a free society. Because if we can’t define the line, we risk either losing our freedoms—or allowing them to be abused in ways that damage the country itself To watch the entire episode: Rumble: YouTube:

Where does free speech end—and real danger begin? In this clip, I explain how the legal standards around speech have evolved in America, and why today’s environment makes it harder than ever to define the line between protected expression and incitement. I walk through how the First Amendment has been interpreted over time—especially with changes introduced by the courts that expanded protections for speech, even when it’s offensive or extreme. But here’s the real issue: not all speech is equal. We already limit certain types—whether it’s commercial speech, explicit content, or harmful material—and yet when it comes to rhetoric that can inflame or incite real-world harm, the standards become murky. I raise the question: what happens when speech crosses into something more dangerous—and who decides? This isn’t about silencing debate—it’s about understanding responsibility in a free society. Because if we can’t define the line, we risk either losing our freedoms—or allowing them to be abused in ways that damage the country itself To watch the entire episode: Rumble: YouTube:

Mark R. Levin

34,769 次观看 • 3 个月前

Young Bob’s speech to the Southampton Patriots was a call to keep marching, to reject mass immigration, and to stand proud for Britain. This is what real patriotism looks like.

Young Bob’s speech to the Southampton Patriots was a call to keep marching, to reject mass immigration, and to stand proud for Britain. This is what real patriotism looks like.

Young Bob

81,600 次观看 • 10 个月前

Listen to this whole speech. This is EXACTLY what Walter did in Fringe when he tapped into a nearly identical dimension and started copying advanced tech from it...except according to this here, it is very real.

Listen to this whole speech. This is EXACTLY what Walter did in Fringe when he tapped into a nearly identical dimension and started copying advanced tech from it...except according to this here, it is very real.

Biff

34,966 次观看 • 6 个月前

You have to watch this video. You know what I took away from it? That Luke is a real one. He doesn’t just say this stuff in his car. We need more vocal outspoken negative speech about our dictator proving we aren’t afraid of using our first amendment right.

You have to watch this video. You know what I took away from it? That Luke is a real one. He doesn’t just say this stuff in his car. We need more vocal outspoken negative speech about our dictator proving we aren’t afraid of using our first amendment right.

MAGA Cult Slayer

19,678 次观看 • 5 个月前

Speech-native models like Moshi sound great and answer fast, but aren’t as smart as text LLMs. In our new paper, MoshiRAG, we show how Moshi can ask for advice from a text LLM or a knowledge base. The tricky part is how to do this in real time without adding latency. 🧵

Speech-native models like Moshi sound great and answer fast, but aren’t as smart as text LLMs. In our new paper, MoshiRAG, we show how Moshi can ask for advice from a text LLM or a knowledge base. The tricky part is how to do this in real time without adding latency. 🧵

kyutai

53,142 次观看 • 4 个月前

When SMEs pay for AI because it’s affordable and useful, that’s when utility becomes real. This CreateWhiz user testimony reflects what MWX is building: real SMEs, real revenue, real adoption powering the ecosystem. CreateWhiz isn’t built to impress AI experts. It’s built so SMEs can create content without learning prompts, hiring extra staff, or increasing costs. Upload → choose → export. That simplicity is why it gets used daily. Every paid AI workflow on MWX generates real revenue. That revenue flows through $MWXT: enabling burns, buybacks, and a growing utility loop. Real SMEs → real usage → real value. That’s how MWXT is powered.

When SMEs pay for AI because it’s affordable and useful, that’s when utility becomes real. This CreateWhiz user testimony reflects what MWX is building: real SMEs, real revenue, real adoption powering the ecosystem. CreateWhiz isn’t built to impress AI experts. It’s built so SMEs can create content without learning prompts, hiring extra staff, or increasing costs. Upload → choose → export. That simplicity is why it gets used daily. Every paid AI workflow on MWX generates real revenue. That revenue flows through $MWXT: enabling burns, buybacks, and a growing utility loop. Real SMEs → real usage → real value. That’s how MWXT is powered.

MWX | Real-Time Alpha AI

39,819 次观看 • 8 个月前