AI in the Real World: Tools I Use and Trust

͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏

Forwarded this email? Subscribe here for more

You’re reading the My News Biz newsletter, which I will be sending you every other Thursday. My goal is to help you and other digital media entrepreneurs to find a viable business model that works for you. If you were forwarded this email, you can sign up here.

AI in the Real World: Tools I Use and Trust

Here’s how AI is already helping me with video, translation, coaching, and troubleshooting.

James Breiner

Jul 3

READ IN APP

You’re reading the Your News Biz newsletter. My goal is to help digital media entrepreneurs find viable business models.

Lately, it seems AI is either here to save us, replace us, or ruin everything. I’ve been thinking about how to help you sort the signal from the noise. This is my third in a series of posts about the new technology.

My two previous posts were:

When Big Tech controls the message
AI in the Attention Economy: Friend, Foe, or Fix?

Some of my favorite AI tools

No. 1. Kapwing has helped me produce short videos for YouTube, Instagram, and TikTok, among others. This is where a lot of young people, 18-35, are getting their news, and my audience has typically been — ahem — somewhat older.

The Reuters Institute’s report “How Young People Consume News: And the Implications for Mainstream Media” pointed out that “Traditional news brands see news as: what you should know. Young audiences see news as: what you should know (to an extent), but also what is useful to know, what is interesting to know, and what is fun to know.” (Emphasis mine — JB).

How to use Kapwing (video tutorial here). This tool uses AI to process long videos into short clips that can be shared to social media. I recently processed a 45-minute video interview of a young journalist recorded on Zoom into 10 clips of 30 seconds each.

Natalie Van Hoozer offered career advice to young journalists in a series of short videos.

Kapwing used some keywords I gave it (journalism, careers, communications, jobs) to output a nearly error-free transcript and 10 clips with headlines. I was able to reword the headlines, correct transcript errors, override the text selected, and produce videos that I uploaded to YouTube. The full playlist has been viewed 170 times, but I believe that will grow since these videos have evergreen content.

Kapwing helped me produce a translation and subtitles of an interview with Ismael Nafria, left, who publishes the newsletter Tendenci@s in Spanish.

Kapwing also did an excellent job of translating and transcribing a Zoom interview I did in Spanish with a journalism colleague. Again, the tool allows me to edit the translation and subtitles; the output is very good but still requires some corrections.

They have a free version, but I use a paid version that gives me priority in the processing queue and faster processing.

My volunteer work. And one more useful Kapwing application. I volunteer as a mentor to small business people with SCORE.org. Like any nonprofit, we need to recruit people and donors, and we need to do it without a big marketing budget.

So at the suggestion of a colleague, I used my iPhone at one of our social events to record a series of interviews of SCORE volunteers about why they do it. Kapwing helped me produce 10 videos of around 30 seconds each. I think my colleagues come off as sincere and credible, perhaps because of the casual, improvised nature of the videos.

Screenshot of three of the interviews I did of SCORE colleagues.

See what you think. The playlist has gotten about 550 views so far.

Leave a comment

No. 2. AI has made language translation much better in the past two years. The large language models have digested much more content to improve their accuracy. All of the tools have different strengths. There is more content available in Spanish (500 million speakers from two dozen countries) than, say, German (which I also follow). That means German translations are often clunky, too literal, and filled with confusion about the subject of a verb (it could be he, she, or it). They need a lot of human intervention.

I’m fluent in Spanish and publish in both English and Spanish, so I can compare the translations I get from different tools and judge their quality. My favorite is Deepl. I use it a lot to save time. It can capture the nuances of my personal writing style (I use a lot of metaphors and idioms) to translate my work into Spanish. Still, it’s not perfect, and I often have to do extra research to make sure an idiom gets a meaningful equivalent.

I’ve also seen Google’s Gemini and Claude get better at translation. I did a comparison test of several of these tools, and there were clear differences in quality. But that was a while ago. These tools can learn from their mistakes.

Pro tip from Jeremy Caplan of Wonder Tools: He doesn’t give a thumbs up or down to an AI response because it gives the tool permission to use your query to train itself.

No. 3. I use ChatGPT to help coach small business owners

As I mentioned above, I work as a volunteer coach of small business people with SCORE. During a typical coaching session (my specialty is digital marketing and social media), I often suggest that my clients do market research. Then I share the screen and let my client ask questions of ChatGPT, such as:

People aren’t responding to my Facebook posts. What should I do?
What price should I charge for my product or service, based on competitors in my immediate vicinity?
Can you craft a business plan for me so I can get a loan?
How many competitors for my service are there in the following zip codes? (Competition is usually fierce for landscapers, home repair and remodeling, beauty salons, food trucks, restaurants, bookkeeping, and child care, among many others.)

The beauty of ChatGPT is that the client, with my help, can ask follow-up questions and get deeper and better research. Obviously, the client has to be cautioned that these answers need to be checked. The tool will sometimes produce outdated information or even invent some in order to satisfy the query.

For deeper recommendations, I recommend technology and management guru Ethan Mollick’s Using AI right now: A quick guide. Below is his summary graphic. To see a larger version, click here.

To see the a larger version of this image, click here.

No. 4. Help troubleshooting technical problems

Whenever I’m having trouble getting a device or an app to do what I want it to do, I no longer use the HELP or SUPPORT buttons that come with that service. I get far better solutions from ChatGPT, Claude, or Perplexity.

Some of my recent issues

I could not get my Apple messaging app to send messages to my contacts using Android phones. ChatGPT walked me through about a dozen different troubleshooting scenarios (the tone of the chatbot’s responses gradually started to sound as frustrated as I was!). Finally, it boiled down to one setting buried deep, deep, deep within the bowels of the iPhone.
My thumb strayed and hit the wrong key on the TV remote, and an annoying robot voice started announcing the names of shows in the screen directory. I consulted the TV’s paper manual, its online help site, replaced the remote’s batteries, restarted the TV several times — nada. Turns out the robot voice is driven by a well hidden setting designed for the blind and visually impaired (a wonderful, thoughtful feature!). I never thought to look at that setting. But that solved it.
Finally, whenever (and it’s often), the HELP solutions of an app or device direct me to a particular button that does not appear on my screen, I go to AI for a solution. It often occurs because of a software update or some Apple-Android incompatibility.

Before we go:

Here’s a little gift for newsrooms that produce newsletters. A local news publisher with a small staff wanted to produce a morning newsletter based on the previous day’s articles without adding staff or increasing the workload. How could he do it? Sophie Culpepper of Nieman Lab tells how Local News Now in Virginia solved the problem.

Páginas