Season's Savings | ₹5/mo for 3 months

₹199/month after first 3 months. Renews automatically. Terms apply.
How to Make AI Write in Your Voice, Protect Privacy and Cut Costs cover art

How to Make AI Write in Your Voice, Protect Privacy and Cut Costs

How to Make AI Write in Your Voice, Protect Privacy and Cut Costs

Listen for free

View show details
About this listen

Can AI write like you without sounding robotic, exposing confidential information, or running up a massive monthly bill? On this episode of Real AI Use Cases Business Owners Roundtable, hosts Elizabeth Gearhart, Ph.D., and Richard Gearhart talk with Arshavir Blackwell, founder of YourVoiceCraft, about local AI models, personalized AI writing, privacy, and the true cost of artificial intelligence.

Arshavir explains how local AI models can be trained on examples of your own writing to produce content that sounds more like you than generic AI-generated text. Unlike hosted AI models such as ChatGPT and Claude, locally run models can offer greater control over your data and avoid per-token API costs, although hardware, setup, and electricity still have costs.

The conversation also explores why intensive AI use can generate surprisingly large bills, how AI pricing is affected by computing and infrastructure expenses, and why data centers consume so much energy and water. Could AI help solve its own energy challenges? And what might happen as AI systems become increasingly capable of designing other AI systems?

If you're a business owner, entrepreneur, marketer, content creator, or AI enthusiast, this episode offers a practical look at the trade-offs between AI convenience, customization, privacy, cost, and sustainability.

TL;DL — Too Long; Didn't Listen

  • AI that sounds like you: How training a local AI model on your writing samples can produce more authentic, personalized content.
  • Lower AI costs: Why heavy ChatGPT and Claude usage can become expensive and how local models may reduce recurring token charges.
  • AI privacy: How running models locally can help keep confidential business information off external AI servers.
  • AI energy consumption: Why data centers require substantial electricity and cooling, and whether AI can help address these challenges.
  • The future of AI: Why the next major advances may come in unpredictable bursts, particularly as AI systems begin designing other AI systems.

Frequently Asked Questions:

How can you train AI to write in your own voice?

According to Arshavir Blackwell, a local AI model can be trained using examples of your existing writing. This helps it reproduce your style, tone, and way of expressing ideas rather than relying solely on prompts to imitate your voice.

What is local AI, and how is it different from ChatGPT or Claude?

Local AI runs on hardware you control, such as your own computer or a private server. Hosted services such as ChatGPT and Claude run on external infrastructure. Local models can offer more control over data and customization, but their performance depends on the model and available hardware.

Can running AI locally reduce AI costs?

Yes, it can reduce or eliminate per-token charges for locally run models. However, hardware, electricity, maintenance, and setup can still cost money. Whether local AI is cheaper overall depends on usage and infrastructure.

Is local AI better for business data privacy?

Local AI can help businesses keep sensitive information within their own infrastructure rather than sending it to an external AI service. Actual privacy depends on how the system is configured, secured, and operated.

Why do AI tools sometimes have such high monthly bills?

AI services may charge according to the number of tokens processed, including the text and other data used in requests and responses. Heavy usage, complex tasks, and AI agents performing repeated tasks can increase costs substantially.

How much energy do AI data centers use?

AI data centers require significant electri...

Chapters
  • (00:00:00) - Introduction: VoiceCraft and Local AI Models
  • (00:00:46) - How Local Models Differ from Frontier Models
  • (00:03:46) - Pricing and Cost of AI Services
  • (00:04:49) - Why AI Tokens Are Subsidized and Expensive
  • (00:06:08) - Data Center Energy and Cooling Challenges
  • (00:09:21) - Future of AI: Next 6 Months to Years Ahead
  • (00:11:25) - AI Self-Designing Systems and Unpredictable Innovation
  • (00:12:35) - Sci-Fi Predictions: Heinlein and Arthur C. Clarke
  • (00:13:19) - Arthur C. Clarke's Genius and Predictive Power
  • (00:14:47) - Closing Jokes: Twilight Zone and Contact Info
adbl_web_anon_alc_button_suppression_t1
No reviews yet