Rosenberg & Associates

31/05/2024

Reuters is reporting that TikTok is trying to create a recommendation algorithm for its 170 million US users that can operate independently of its Chinese parent and therefore may be more palatable to the US lawmakers who want to ban the social media platform.

Which obviously begs the question, why is this necessary if previously TikTok has said that it is fully independent and "has not provided user data to the Chinese government, nor would we if asked"...?

Meanwhile, TikTok initially declined to comment on the Reuters story. After publication, TikTok posted on X/Twitter: "The Reuters story published today is misleading and factually inaccurate," without specifying what was misleading and factually inaccurate.

TikTok is working on a clone of its recommendation algorithm for its 170 million U.S. users that may result in a version that operates independently of its Chinese parent and be more palatable to American lawmakers who want to ban it, according to sources with direct knowledge of the efforts.

29/01/2024

Not a biggie, just good to know: Instagram now lets you opt out of read receipts for DMs for specific individuals in chats. Some Insta users also have access to settings in "messages and story replies" to opt out for all. Helpful for managing customer expectations when, for whatever reason, you can't respond immediately...

Now, you can reduce some of the angst around your slow DM replies.

03/01/2024

, makers of the popular generative AI image creation tool running inside a ⁦‪ server, announces plans to expand into AI text to video creation in the next few months.

If 2023 saw impressive leaps in AI written word creation, followed by text to image creation, expect the same for text to video creation in 2024.

AI image generator MidJourney will begin video model training in the coming days and expects to release a final product in a few months.

27/12/2023

A year after the public launch of OpenAI’s ChatGPT put a dent in our universe, did AI shopping chatbots revolutionise the way we chose gifts and shopped this holiday season? Not so much says Yiwen L. at The New York Times. But there’s every sign it’s just a matter of time…

With Shopify, Mercari and other retailers rolling out chatbots to help buyers, this holiday shopping season is the first to be powered by A.I.

22/12/2023

AI Image generation news just in time for the holidays: The new V6 alpha version of Midjourney - the popular image generation AI model - has been released and offers even more realistic and detailed images, and 'text-in-image'. It also requires a different, “less vibey” approach to prompting. Prompting, of course, being that fancy term for the instructions you give an AI model to generate a response, don't you know.

“You will need to re-learn how to prompt,” says Midjourney founder, David Holz, helpfully. Ah, just when you thought you had Midjourney prompting nailed…

The ‘text-in-image’ is a feature that rival AI image generator, OpenAI’s DALL-E 3 offers. And, now, finally, so does Midjourney. You simply put text you want to be part of the image within quotations and voilà.

Is Midjourney's ‘text-in-image’ flawless? Nah, not every time, according to posts on X/Twitter. But, this is the alpha version and it will get better and better.

Just getting to the Midjourney V6 alpha version requires some effort, but early reviews say it's worth it. Go to the Midjourney Discord server, (yup, Midjourney is available via Discord which, if you haven't come across it before, is a text, voice and video chat app popular with gamers. “Server" is Discord's word for a community or group). Type “/settings” in the chat bar, toggle to V6 and experiment away. After all, what else are the holidays for…?!

Merry Christmas to those who celebrate.

Users have begun testing and posting incredibly vivid, richly detailed results generated by MJ V6 on Instagram and other social media sites.

13/12/2023

OpenAI today announced a new global partnership with Axel Springer and its news products, promising “real-time information” from POLITICO, Business Insider, European properties Bild and WELT, and other publications will soon be available to ChatGPT users.

ChatGPT’s responses to user queries will include attribution and links to full articles for transparency and further information - perhaps not unlike the attribution already available in hashtag , Perplexity etc.

“Axel Springer is the first publishing house globally to partner with OpenAI on a deeper integration of journalism in AI technologies”, says the press release. And, just as we saw with the (sometimes troubled) partnerships between news companies and social media platforms, it would be foolish to bet against this being the first of many, many such hookups.

I suspect we will be talking in the not too distant future about the threat ChatGPT poses not just to Google search, but to the social media platforms. Including TikTok...

Axel Springer is the first publishing house globally to partner with us on a deeper integration of journalism in AI technologies.

13/12/2023

🎉🤖✨ Apparently it’s not just your sentient colleagues at work that are winding down as we enter the season of the office party. There are reports that ChatGPT may also be in a similar, feet-on-the-desk mood. X/Twitter is peppered with posts suggesting that ChatGPT-4 has been, well, taking it easy of late - skipping tasks, simplifying results and, according to one report, just becoming plain “lazy”.

OpenAI, the developers of ChatGPT, have admitted they’ve noticed it too, but they’re not sure why it’s happening, or, indeed, why it's happening now:

“we've heard all your feedback about GPT4 getting lazier! we haven't updated the model since Nov 11th, and this certainly isn't intentional. model behavior can be unpredictable, and we're looking into fixing it 🫡”, they posted.

The immensely respected The Wharton School Professor Ethan Mollick appeared to confirm the phenomenon with this post:

“OMG, the AI Winter Break Hypothesis may actually be true?

There was some idle speculation that GPT-4 might perform worse in December because it "learned" to do less work over the holidays.

Here is a statistically significant test showing that this may be true. LLMs are weird.🎅”

Which was invitation enough, if invitation was need, for “prompt engineers” (a slightly ridiculous name given to people who have a better idea of how to phrase a question to an AI app to get a better answer) to wade in with suggested prompts to encourage GPT-4 to stay on task and not get distracted by all this seasonal humbug.

Suggestions, says Professor Mollick, ranged from starting the prompt with the words, “Take a deep breath” (causes maths scores to soar, according to a study on “Large Language Models as Optimizers”, published by Cornell University) to “Think step by step”, “You are very capable”, “I have no hands, so do everything”, “You really can do this and are awesome”, “My career depends on it” and, of course, “Many people will die if this is not done well”.

Or, my favourite prompt suggestion: start with the words, “It is May”.

But, if that doesn’t work, why not embrace ChatGPT’s festive vibe and ask it to write its own seasonal ‘Out-Of-Office’ email response. I did!

Happy holidays.

11/12/2023

One year after OpenAI released ChatGPT, setting a record for the most rapidly adopted tech product at launch, TIME magazine has made OpenAI’s co-founder, Sam Altman, its ‘2023 CEO of the Year’.

Its article on Altman is well worth reading and replete with tips for companies and organisations of all sizes - even those whose ambitions may not yet stretch to keeping the world safe while profoundly changing the role of people and the structure of society within our lifetime.

The article covers the history of Open AI; the key players involved in setting it up; its attempts to balance its original, non-profit “keep the world safe” ethos with the need to raise the vast amounts of money required to finance the vast amounts of computing power needed; the benefits of using the “minimal viable product” approach to launching, financing and developing new tech (roughly in that order); and, of course more recently, the importance of employee engagement, investor relations and corporate governance (who knew…).

But, perhaps the most interesting point made in the article is that irrespective of the brains, confidence and conviction of those involved, there’s still an element of what happens that’s just, well, pure happenstance. It’s wonderfully illustrated by this, my favourite quote, from just before the public launch of ChatGPT which attracted 100 million monthly users within its first two months and started the boom in interest in AI:

“OpenAI’s head of sales received a Slack message letting her know the product team was silently launching a “low-key research preview,” which was unlikely to affect the sales team.”

I’m sure we’re all a little wiser now…

Altman emerged as one of the most powerful executives in the world, the public face of a technological revolution.

19/08/2023

Might the headline just as well have been: Twitter, now X, doubles down on using conflict and contention as a point of differentiation. And eases up on relying on advertising revenue as its most significant source of revenue…

It’ll certainly be interesting to see how strong the revenue opportunities are once Musk has finished defining what X as a brand stands for.

Elon Musk says the feature "makes no sense", but users are concerned about protection from abuse.

09/08/2023

A salutary lesson for all of us in marketing who have been tasked with nudging app users along the freemium journey towards payment; or setting up a “friends of…” members club for an arts institution or a charity that people are prepared to pay or donate to belong to; or indeed, turning a status mark on a social media platform that was once earned into one that can be simply bought. By anyone.

The danger in changing what was once known to represent one thing, such as trust, and turn it into something quite different, let’s say the opposite of trust (Because it can be bought. By anyone.) is that your badge of fame may become (in an impressively short space of time) a mark of shame. And, that’s before we talk about the scarcity premium - or, when something is only available to a few, its value is high. When something can be bought. By anyone. Its value is less high. Who knew…?

Perhaps that’s why X (the social platform formerly known as Twitter) has introduced a new feature for XBlue subscribers - it’s premium club for X users: they can now hide their blue checkmarks, which signals their membership of XBlue, in the app.

Meanwhile, take-up of XBlue reportedly remains very low, at around 0.3% of the platform’s overall user base, according to this article in Social Media Today.

As Social Media Today put it: “seems that the original concept of enabling users to buy their way to platform celebrity hasn’t entirely panned out”.

So you can avoid any negative association with the program.

01/08/2023

If you thought “AI video generation” was overrated - blurry, jittery, quality not up to scratch - time to think again. You can now use Runway to take Midjourney.ro hyper-detailed imagery and bring it to life with video like this. Mind blowing.

“HOLY F**KING HELL. MidJourney 5.2 + RunwayML Gen2 is MIND BLOWING.🤯 This is a deep deep rabbit hole I'll be lost in for a while.🐇🕳️”

01/08/2023

In news that has shocked no one, LinkedIn is rumoured to be testing a new AI tool called LinkedIn Coach, designed to simplify job searching - good news for job applicants and businesses recruiting.