Almost Timely News, 6 February 2022: Two AI Advances, Dedication, Content Strategy (2/6) :: View in Browser

Almost Timely News

Learn how to improve your Twitter analytics with my free hands-on masterclass »

Watch This Newsletter

Almost Timely News 6 February 2022
Watch this video on YouTube.

Click here for the video version of this newsletter on YouTube »

Click here for an MP3 audio only version »

What’s On My Mind: Two Advancements in Applied AI

Let’s talk about two recent advancements in AI that have impressed me greatly in recent days. First, let’s talk about natural language generation. For about a year and a half, I’ve been using the GPT-J-6B model from Eleuthera AI for generating text. It’s a powerful model that creates coherent, readable text given a prompt. This past week, Eleuthera released their newest model, GPT-NeoX-20B, a model that’s more than 3 times the size and capability than the previous model. It does very well on a bunch of technical benchmarks, but the real question is, how well does it write compared to prior models?

The answer is – really well. Really, really well. It’s still constrained by the text it’s trained on, so it will never generate works that are above and beyond that training data, but it’s getting much better at generating content similar to what’s in the training dataset. Press releases? It has those down now. News articles? Corporate boilerplate? All these types of content are now well within its capabilities.

Let’s look at a quick example, a press release prompt. I’m going to make up one entirely here and we’ll see what the machines write.

Prompt:


FOR IMMEDIATE RELEASE

Boston, MA, February 6, 2022– Trust Insights Inc., a management consulting and analytics firm based in Boston, is pleased to announce its newest product line, the Marketing Analysis Pack. The Marketing Analysis Pack provides forward-thinking companies with key analytics and insights about the top of their marketing funnel to more quickly make data-driven decisions.

Katie Robbert, CEO of Trust Insights, remarked, “We are


Here’s what the previous model, the J-6B model, came up with. I’ve trimmed it for space’s sake.


Katie Robbert, CEO of Trust Insights, remarked, “We are excited to launch the Marketing Analysis Pack because we are seeing an increase in demand for more actionable insights to make marketing decisions, especially in the context of big data.” She continued, “The Marketing Analysis Pack can help companies build an intelligent marketing strategy to more efficiently capture leads and improve conversion.”

Trust Insights’ Marketing Analysis Pack is comprised of four modules:

Lead Management Module

The Lead Management Module provides insight into your lead management strategies, how they are working, and how you can improve them. You will learn how to predict, quantify, and target your leads better, and this will allow you to optimize your marketing spend and cut costs while also improving lead generation and conversion.


That’s not bad, right?

Let’s see what the new NeoX-20B model came up with.


Katie Robbert, CEO of Trust Insights, remarked, “We are excited about the opportunity to provide marketing teams with a toolset that provides insights and information they need to be successful in this increasingly competitivemarket.”

According to a 2019 CMO survey, 86% of marketers plan to increase budgets on paid search and social media ads in 2020, but only 10% of marketing teams actually have the analytics necessary to make informed business decisions. In fact , only 19% of marketing departments use data to optimize their campaigns and budgets. As a result, marketers are making the wrong decisions about their marketing programs.

“The Marketing Analysis Pack provides the analytics needed for companies to make data-driven marketing and sales decisions,” explained Robbert. “The pack is designed with the needs of B2C marketers in mind, but it is adaptable enough to support the needs of B2B marketers.”

The Marketing Analysis Pack includes a dashboard that allows users to track and view data from Google, Facebook, and other platforms, including ad performance, brand lift, audience insights, and social media sentiment. Additionally, users can see the most active keywords, ads, and social media posts, and can view social media sentiment, and even create a list of influencers that are the most engaged on each social network.


Both models spit out usable content, especially for a press release. The 6B model is clearly cribbing from some other product or service, though I couldn’t tell you what. The 20B model appears to be writing more original text (though it’s still cribbing from somewhere, as evidenced by the incorrect dates), and really does a great job capturing the tone of a press release and the CEO quote.

Impressive, right? It’s not good enough to write completely original content or start with a minimal prompt, but it’s definitely good enough for saving you a lot of time on first drafts – especially the newer model. That said, neither entry is factually correct; if we were going to use this, we’d have to rewrite those portions for what the analysis pack actually is. But on balance, this is terrific writing for a machine.

That’s the first tool of note. The second is a fully commercial, in-market tool called DeepL. In my work and experimentation with newer social networks like Discord, one of the persistent issues I have is that there are a lot of people on the planet who don’t speak my language, and I don’t speak theirs. Over the years, tools like Google Translate have evolved our translation skills so that we can be understood by others in their language. However, there’s always room for improvement – and that’s what DeepL is, an improvement. It supports far fewer languages than Google does, but it does a better job of converting that language into something not only readable, but understandable.

Let’s take this short passage of fiction in Danish by author Line Jensen:

Men du skal vide, at alt jeg sagde til dig, var det jeg mente. Jeg var altid ærlig overfor dig – måske endda mere, end jeg var overfor mig selv til tider. Og jeg ville bare ønske, at du havde indset, at du ikke kunne ‘fikse’ mig, fordi jeg var aldrig rigtig helt ødelagt. Men du kan heller ikke omgås mig, bare fordi, at du er bange for at være alene – det er ikke fair overfor hverken mig eller dig. Du fortjente ikke at kende mig, på den måde, som jeg tillod. Du fortjente ikke at kende mig overhovedet. Jeg er bare ked af, at jeg ikke opdagede tidligere, at jeg kun eksisterede i din verden, når du havde brug for noget. Måske, er det bare mig, som tænker for meget, og er for dum til at finde det rigtige svar, men… Hvorfor brugte du så meget tid på mig, og med mig, hvis ikke du havde intentioner om at blive i mit liv? Hvis din ide af ‘farvel’, var at ignorere mig, indtil jeg gav op, så tillykke. Du har vundet, fordi jeg kan ikke længere holde til at kæmpe for dig. Du ignorere mig, som om det ikke betyder noget, men når jeg så meget som forsøger at have en normal dag, uden at snakke med dig, så kan jeg aldrig holde op med at tænke på dig, og det er ret så frustrerende. Mine følelser raser lige for tiden mellem fortvivlelse, vrede, uretfærdighed, hævnlyst, men mest af alt; hjertesorg.

I chose fiction, and this passage in particular, because it’s much more colloquial than more formal business writing. This character’s speech is very much like how a regular person would speak or write. Let’s see how the machines did.

Here’s Google Translate:

But you must know that everything I said to you was what I meant. I was always honest with you – maybe even more than I was with myself at times. And I just wish you had realized that you could not ‘fix’ me because I was never really completely ruined. But you can not associate with me either, just because you’re afraid of being alone – it’s not fair to either me or you. You did not deserve to know me, in the way I allowed. You did not deserve to know me at all. I’m just sorry I did not discover earlier that I only existed in your world when you needed something. Maybe, it’s just me who thinks too much, and is too stupid to find the right answer, but… Why did you spend so much time on me, and with me, if you had no intentions of staying in my life? If your idea of ‘goodbye’ was to ignore me until I gave up, then congratulations. You have won because I can no longer stand to fight for you. You ignore me as if it does not matter, but when I’m so much as trying to have a normal day, without talking to you, I can never stop thinking about you, and that’s pretty frustrating. My emotions are currently raging between despair, anger, injustice, vengeance, but most of all; heartache.

And here’s DeepL:

But I want you to know that everything I said to you was what I meant. I was always honest with you – perhaps even more than I was with myself at times. And I just wish you had realized that you couldn’t ‘fix’ me, because I was never really all that broken. But you can’t be around me either just because you’re afraid to be alone – that’s not fair to me or to you. You didn’t deserve to know me the way I allowed you to. You didn’t deserve to know me at all. I’m just sorry I didn’t realize sooner that I only existed in your world when you needed something. Maybe, it’s just me thinking too much, and being too stupid to find the right answer, but… Why did you spend so much time on me, and with me, if you had no intentions of staying in my life? If your idea of ‘goodbye’ was to ignore me until I gave up, then congratulations. You’ve won, because I can’t stand fighting for you anymore. You ignore me like it doesn’t matter, but when I so much as try to have a normal day without talking to you, I can never stop thinking about you, and it’s pretty frustrating. My emotions are raging right now between despair, anger, injustice, vindictiveness, but most of all; heartbreak.

We see that DeepL produces output which is less stilted, less overly-formal. It’s more how the author’s character meant to sound and probably did sound in the original.

So What’s the Takeaway Here?

These two language tasks – language creation and language translation – are tasks we have to do often, depending on our line of work. In the past, both tasks had been attemped by AI with… well, not great results. A few years ago, advances in neural networks and deep learning brought those tasks from not great to mediocre to not bad, and now we’re at a point where they’re pretty good. You don’t need to be a data scientist to realize the trajectory of improvement here is substantial – and in a relatively short period of time.

This means two things for you. First, you should be evaluating AI solutions for specific, narrow tasks in your organization and deploying it to fulfill those tasks as rapidly as possible. If, for example, you crank out press releases, AI should be doing all your first drafts now. If you make captions in videos for different language markets, AI should be doing all the translations of those captions. I did a video recently for a friend in French and they said the translation was spot on, helping me be more understandable in their language. For fun, I’ve put Spanish captions into this video as an example. Is the language translation perfect? No. Is it’ usable? Absolutely.

Second, these tools open doors. If you have friends you’ve met online or at in-person events and your language isn’t their native language, you now have a way to communicate with them easily. It won’t be via voice in real-time, but you can absolutely hold a chat conversation in many common languages without needing to speak it and still be understood.

If you’re working on developing a product description and you have no idea what to write about your product, you now have tools to do the first draft. The Marketing Analysis Pack I talked about above does exist, but we haven’t gotten around to building the marketing for it yet. Those fake press releases that AI generated have given me a good first draft, a good starting point for how we might want to actually talk about them.

AI is a superpower of sorts. With nothing more than apps on your phone or desktop computer, you can do more, create more, learn more, communicate more. You can exceed your own limits as a person. I can’t read Danish, but AI can. I can’t write an entire press release in 30 seconds, but AI can. With these powers, we can dramatically increase what we’re capable of as people.

Share With a Friend or Colleague

If you enjoy this newsletter and want to share it with a friend/colleague, please do. Send this URL to your friend/colleague:

https://www.christopherspenn.com/newsletter

ICYMI: In Case You Missed it

If I had to pick one thing for you to take a look at, it’s the Mind Readings piece on dedication and consistency – and what we’re telling our audiences through our actions as marketers and content producers.

Skill Up With Free Classes

These are just a few of the free classes I have available over at the Trust Insights website that you can take.

Get Back to Work!

Folks who post jobs in the free Analytics for Marketers Slack community may have those jobs shared here, too. If you’re looking for work, check out these five most recent open positions, and check out the Slack group for the comprehensive list.

Ad: How to Prove the ROI of your Marketing Agency

I put together a brand new talk on how agencies could use data-driven marketing as a way to showcase their value and real results they obtain. Fundamentally, agencies need to take five steps to make this journey:

  1. Become data-driven. Making decisions with data allows you to act faster and make better decisions when done right.
  2. Be crystal clear about KPIs. What’s a KPI? It’s the number you get your bonus for (or fired for).
  3. Build an agency cookbook. Cookbooks set apart good agencies from great ones.
  4. Use data to become proactive. Impress clients by being more proactive and pushing them.
  5. Squeeze all the juice from your tools. You probably don’t need to buy more tools.

An agency which takes these steps becomes more and more valuable to its clients. For folks on the client side, these are the things you should expect of your agencies, things you should ask for when agencies are pitching you. Agencies not doing these things will not serve you as well as they could. There’s obviously a lot more detail, so go ahead and watch the talk now.

Watch the talk now by filling out this form »

What I’m Reading: Your Stuff

Let’s look at the most interesting content from around the web on topics you care about, some of which you might have even written.

Social Media Marketing

Media and Content

SEO, Google, and Paid Media

Advertisement: Supermetrics for Google Data Studio

Google Data Studio is an incredibly powerful tool for bringing your data into one place for analysis. Out of the box, it serves up Google ecosystem data that marketers need – Google Analytics, Google Search Console, YouTube, Google Ads.

But what about… Facebook data? LinkedIn data? SEO data? Data from your CRM or marketing automation software? That’s where Google Data Studio has some gaps.

Fill those gaps with the Supermetrics Integration for Google Data Studio with over 70 different supported integrations like most social networks, CallRail, Adobe Analytics, Bing Ads, Pinterest, Stripe, Semrush, Salesforce, and so many others. Get ALL your marketing data in one place for easy reporting.

Click here for a free trial of Supermetrics for Google Data Studio »

Tools, Machine Learning, and AI

Analytics, Stats, and Data Science

All Things IBM

Advertisement: MarketMuse for SEO

How do you create content at scale? You could hire dozens of copywriters and SEO specialists – or you could engage an army of machines to jumpstart the process. That’s what MarketMuse is – an army of AI content creators at your fingertips. Research keywords based on topics instead of narrow phrases, have machines write your first drafts, and boost your SEO results immediately.

Try it free for 7 days and see what having your own SEO robot army can do for you »

How to Stay in Touch

Let’s make sure we’re connected in the places it suits you best. Here’s where you can find different content:

Required Disclosures

Events with links have purchased sponsorships in this newsletter and as a result, I receive direct financial compensation for promoting them.

Advertisements in this newsletter have paid to be promoted, and as a result, I receive direct financial compensation for promoting them.

My company, Trust Insights, maintains business partnerships with companies including, but not limited to, IBM, Cisco Systems, Amazon, Talkwalker, MarketingProfs, MarketMuse, Agorapulse, Hubspot, Informa, Demandbase, The Marketing AI Institute, and others. While links shared from partners are not explicit endorsements, nor do they directly financially benefit Trust Insights, a commercial relationship exists for which Trust Insights may receive indirect financial benefit, and thus I may receive indirect financial benefit from them as well.

Thank You!

Thanks for subscribing and reading this far. I appreciate it. As always, thank you for your support, your attention, and your kindness.

See you next week,

Christopher S. Penn


You might also enjoy:


Want to read more like this from Christopher Penn? Get updates here:

subscribe to my newsletter here


AI for Marketers Book
Get your copy of AI For Marketers

Analytics for Marketers Discussion Group
Join my Analytics for Marketers Slack Group!


Pin It on Pinterest

Shares
Share This