 |
|
|
|||||
For as long as weve imagined the future, weve imagined computers that talk with humans. From the calm, ever-listening computer in Star Trek to J.A.R.V.I.S. in Iron Man, voice-enabled AI has been the centerpiece of sci-fi and a symbol of technological advancement. Well, that future is now. And voice AI is in the middle of a gold rush. Voice AI interactions have evolved from clunky text-to-speech tools with voices that sound like robots to new conversational voice AI technology that resembles human speech so closely its eerie. We can talk to ChatGPT and get voice responses that feel thoughtful, funny, and authentic. Googles AI search can now talk to you while searching the web and answer questions like a well-briefed assistant. These voicebots dont just talk, they converse. They demonstrate that they actually understand what were saying while closely mimicking real spoken communication with pauses, inflection, emotion, context, and tone. And this is only the beginning. Without a doubt, voice is AIs next frontier. But its progress depends on the quality and integrity of the voice data on which its trained. The real gold? Voice data Whats powering this new generation of voice AI isnt just better codeits voice data on which voice models are trained. More specifically, its massive datasets of high quality and diverse human voices, representing the range of human speech in all its complexityacross languages, dialects, vocabulary, patterns, emotions, inflections, and context. Now that the industry sees where AI is headed, its understanding the mission-critical value of voice data, and everyone wants access to this data. Tech giants and startups are scrambling to collect, license, or build it from scratch. Everyone wants to create the next, most lifelike talking AI, and they need the voice data to fuel it. This is the voice data gold rush. But just like the original gold rushes of the 1800s, the current frenzy comes with risk and consequence. If you dont have permission, its stealing I firmly believe that to build voice AI the right way, technically and ethically, the data training your voice AI models needs to satisfy three criteria. The data must be High quality: Clean, extremely high-fidelity human voice recordings that are free from background noise or distortion, represent diverse voices and speech patterns, and offering rich emotional and linguistic content. High volume: Enough data to meaningfully train a model. High integrity: Ethically-sourced with clear licenses and proper consent for use in AI training. Many existing datasets can meet one or two of these requirements. Getting data that hits all three is the hard part. Dont take shortcuts I dont hear many companies talking about how theyre building AI ethically, or clearly stating the sources or permissions behind the data used to build their voice AI. Yes, theyre able to move fast. Many voice AI startups go to market within months. But when theyre able to produce life-like voices that quickly and with very limited capital, I cant help but wonder: Where did all their training data come from? To save time and cut costs, companies are taking shortcuts by scraping audio off the internet, relying on datasets with murky or unknown ownership, or using data thats licensed for AI training, but fails to meet the quality standards needed to train convincing voice models. This is the fools gold of AI: data that looks shiny, but cant stand up to legal scrutiny or meet the appropriate quality standards. The reality is that voice AI is only as good as the data its trained on. And if youre building a voice model meant to reach millions of users, the stakes are high. Your data needs to be clean, consented, licensed, and diverse. Just look at the headlines: AI voiceover company stole voices of actors, New York lawsuit claims. Companies are being called out and sued for cloning and using voices without permission. When you take the unconsented route, youre not just risking a PR headache; you open the door to lawsuits, reputational damage, and most importantly, you risk a major loss in customer trust. Build AI that lasts Were entering a new era of human-to-computer interaction, one where voice is the default interface. AI that talks will soon become the standard way we shop, learn, search, work, and even forge relationships. But for that future to be truly useful, human, and trustworthy, we need to build it on the right foundation. Were still relatively early in the generative AI boom, and navigating the legal landscape around training data rights and licenses is complex. If theres one thing we know for sure, any lasting, successful AI voice product will rely on quality data obtained the right way. The gold rush is here. The smart players arent just chasing shiny things. Theyre building voices that last. Jay OConnor is CEO of Voices.com.
Category:
E-Commerce
Its no secret that the most exciting companies are often run by their founders. Whats less obvious is that the rare opportunities to become CEOespecially of fast-growing tech companiestend to go to people who have been acting like founders all along. Founders often bring the rare combination of vision, grit, and salesmanship that can take a company from idea to juggernaut. As a result, founder-run companies tend to dominate the tech space. In fact, founders remain CEOs at 88% of B2B software companies at IPO, according to recently published SaaStr analysis of 65 major IPOs in this space. I remember being very intrigued during the 2017 Uber CEO search, precisely because of this dynamic. For a company as exciting and transformative as Uber, it was quite rare to have an open CEO role. And for that reason, candidates who were CEOs of much larger revenue scale companies were rumored to be considering the role, including CEOs of companies like General Electric and Hewlett Packard. Dara Khosrowshahi was chosen for the role, and it was no accident that he had trained so closely with IAC founder, Barry Diller. I think most would agree that Khosrowshahi, despite having never been a founder, has been excellent at Ubers helm (and the stock price shows it). The same is true of nonfounder CEOs Satya Nadella of Microsoft, Nikesh Arora of Palo Alto Networks, and so many otherslike Greg Peters, who rose through the ranks at Netflix and became co-CEO in 2023. Show up for the company In early 2024, I started getting calls to interview for CEO roles, including one at Figure, a fast-growing blockchain and fintech company. The founder was considering moving into an executive chairman role. Instinctively, I knew this was a big opportunity. I had worked for its founder, Mike Cagney, before and my fintech experience had shown me the enormous opportunities in solving capital market inefficiencies with blockchain technology. I think a big reason why I had these opportunities is that I had been operating like a founder rather than an executive for years. I remember distinctly when I was the VP of finance at SoFi, but was asked to turn around the mortgage division. A number of people advised me against it: The role was very specific (not every company has a mortgage division, but every company has a finance department) and I had to move out of San Francisco to an operations center far outside the city. I chose to do it, because that was what the company needed. I optimized for the companys needs rather than what I thought was the career-optimizing decision. Ultimately, optimizing for the company was the career-optimizing decision too, as I was later promoted to chief revenue officer. However, whenever I share that anecdotewhether talking with mentees or employees about them making similar moves, I have been almost universally rejected. It can be unnatural to optimize for companies. But for a founder, its second nature. Volunteer for opportunities That lesson came back again at Brex, which I joined following SoFi. Our initial public product launch was highly anticipated internally, as we had worked for over a year to build the product, and were keeping our $50 million in capital raises in stealth. Three months before the scheduled launch, we unexpectedly lost our marketing head. I knew nothing about marketing at the time, but I had witnessed SoFis marketing success firsthand. I was the CFO, but I knew we had to have a marketing launch, so I stepped up, offering to run marketing and the launch. This was a high stakes launch, and I had no experience. It was tough, with a lot of ups and downs. But ultimately it was not only hugely rewarding for me, but also a very successful marketing launch. In looking back at my experiences, and of people I admire, there are a few common themes: Optimize for the company, not your resume. A founder mentality does what needs to be done for the companys success, whereas an executive mentality can view it as presenting more downside than upside. The best opportunities dont involve applications Rather, theyre created by those who think and act like founders long before theyre given the CEO title. They almost always come from inside organizations or via referral. Get comfortable with being uncomfortable. Whether its stepping into unchartered territory or taking on that role that looks like a headacheit might just be your best move yet. The best CEO jobs may seem unavailable or out of reach, but if you act like a founder today, you might just be asked to fill one tomorrow. Michael Tannenbaum is CEO of Figure.
Category:
E-Commerce
TJX Companies has announced the grand opening of six new T.J. Maxx stores later this month. The off-price retailer, which carries major brands at discounted rates, has over 5,000 stores across nine countriesin the United States, it also operates Marshalls, HomeGoods, HomeSense, and Sierra. Currently, there are more than 1,330 T.J. Maxx stores in the U.S. This news comes a little over a year after CEO Ernie Herrman announced the companys goal of adding 1,300 stores to its global portfolio of locations during a quarterly earnings call. T.J. Maxx enthusiasts, read onhere’s the full list of new locations coming soon, as well as the opening date for each. Glen Allen, Virginia Address: 9850 Brook Road, Glen Allen, Virginia Opening date: August 10 Clinton, Connecticut Address: 215 East Main St., Clinton, Connecticut Opening date: August 21 Dickinson, North Dakota Address: 446 18th St. West, Dickinson, North Dakota Opening date: August 24 Saratoga Springs, Utah Address: 1233 North Lake Drive, Suite E, Saratoga Springs, Utah Opening date: August 24 Tooele, Utah Address: 945 North Main St., Suite F, Tooele, Utah Opening date: August 24 Washington D.C. Address: 5300-B Wisconsin Ave. NW, District of Columbia Opening date: August 28 Despite consumers and department stores facing retail uncertainty as tariffs loom, off-price retailers may be at an advantage as they typically receive their goods after supplying brands have already paid the price of tariffs. TJX Companies has seen a steady stock increase over the past five years. The stock price (NYSE: TJX) dipped slightly today before beginning a steady increase this afternoon.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||