You can still visit Lycos today.

Here’s some history from the company’s current about page:

“Lycos is a search engine which originally started as a research project at Carnegie Mellon University and officially launched in 1995. This is before most people even knew what a ‘search engine’ was. The company quickly grew and started providing search results both in the US and in many other countries which it continues to do to this day, effectively establishing a global search brand.”

– Lycos Company Overview

The search engine even does a pretty good job of returning “Hamsterdam SEO news recaps” results:

Now, it’s about time for our introduction of the week!

A look into the future of AI agents driven by Gemini models and vector search …

---

Introduction to week 53: “agent of change”

When I signed up to watch Google Cloud Next 2024, I planned on having it run in the background while I worked.

About five minutes into the first keynote, I stopped everything I was doing. I was mesmerized.

I’m not a flashy guy.

I grew up in a small town in Montana, where I saw people working with tools under the hoods of “beater” trucks more often than using tech in offices.

I sort of like the days when Steve Jobs sat on a stool, sipped a water, and did intellectual battle with the crowd.

That said, the presentation at Google Cloud Next 2024 was pretty amazing. I say this not only for the ability of presenters to type under pressure — there were a few typos, which were relatable human moments amid the polished skits — but the use cases for the AI tech were often next level.

Here’s a Google Cloud article showing a list of real-world use cases for generative AI. (It kind of reminds me of my AI for SEO blog post, which will soon need some major refreshing.)

I don’t see using generative AI to write marketing copy or create stock imagery as aligning with where we’re headed, based on Gen Z’s appeal for TikTok-style relatability in content.

(There’s a story on their search habits in the recap below.)

But there’s a difference between flash (surface level) and innovation (substantive).

The Google Cloud Next presentation had many innovations worth talking about.

One big theme was AI agents.

There were two demos of ecommerce applications for AI that caught my attention and highlight the future of what SEO and digital marketing could be.

The first was a demonstration of a “customer agent” (which you can watch here).

Here was the general workflow:

• The presenter saw a shirt they liked in a music video.
• They visited an online store that has a customer agent that leverages Gemini (AI) and vector search.
• They described the shirt and dropped the video link for context, also asking for prices and where to buy the shirt.
• Using multimodal reasoning, the customer agent turned the results into a searchable format and found the shirt.

• After the presenter added the shirt to a cart for in-store pickup, the customer agent alerted them to low inventory and suggested calling the store.
• The customer agent then engaged a chat on the presenter’s phone and allowed them to add more items and purchase with a card on file.

Pretty impressive.

Of course, why did the presenter choose that video to start with, or visit that store?

If this were an actual shopper, there would be a lot of marketing influences that SEO can play a role in beforehand. Not to mention the data gleaned afterward.

Speaking of which …

The second demo was a “data agent” (which you can watch here).

Here was the general workflow:

• The presenter used a custom application built with BigQuery, Looker, and Vertex AI.
• They got a notification from a data agent that sales were up.
• They looked at real-time enterprise data from BigQuery to surface anomalies and trends.

• They queried the data using natural language and asked for a heatmap.
• The Looker semantic layer grounded the AI data in specific business data.
• The agent looked into projected supply and demand levels over the next 3 months using a future-looking forecast from machine learning and BigQuery data.

• The agent used multimodal data (like product descriptions, pricing (structured data), and images (vector embeddings)) to find alternative products during a time of inventory shortage.

• Using a Workspace integration with Looker, the presenter then shared the findings in a summary over chat along with a sheet of data.

On that note, “context switching” was another common theme.

In the SEO world, we can think of this like eliminating pogosticking among results for a query.

After all, a search engine’s goal is to help users accomplish everything in one click, like getting an AI summary and then selecting the ideal web citation with no added effort — a “qualified” click.

The same applies to the use of AI agents, where the goal is to accomplish everything in one platform interface.

As for how those data queries are possible, “semantic” was another big phrase used, and relatedly, so was “vector,” an embedding of text or other data used for semantic search.

In fact, a day before the keynote, Google Cloud released a YouTube video on vector search and embeddings.

That timing likely wasn’t coincidental.

Keyword-based search still matters, but it’s more accurate to think of a hybrid search model, with lexical, semantic (or vector), and also generative AI.

Google Search is also being used to ground these AI agents and chatbots in more factual information through a RAG (retrieval augmented generation) model.

As Google’s CEO shared in a presentation recently at Stanford, AI will improve Search.

If the presentations at Google Cloud Next were any indication, we’ll have some notable advancements at Google IO next month.

These will likely reaffirm a more personalized and multimodal Search experience, where AI technologies and interfaces are laced throughout, and less “context switching” is required.

As for the role of the SEO in this equation, it probably won’t be adding keywords to pages or chasing backlinks so much, but rather strategically advising businesses on how to get the most value out of their opportunities for cumulative organic visibility throughout user journeys.

How we fully measure that impact is TBD.

Lastly, for a bit of housekeeping, I said last week the March 2024 core update rollout would likely end by now. It hasn’t.

But to complement our AI theme here, if we want to better understand these ranking system updates, one concept worth learning more about is backpropagation.

This is a fundamental algorithm related to machine learning.

As to why it’s relevant …

Gary Illyes from Google responded on LinkedIn this week about how core updates are like adjusting cooking ingredients in a recipe, such as salt or MSG amounts, which can “radically change the result”:

ICYMI: Google says again that indexing and algorithm updates are independent https://t.co/s8mHXOoJ1O pic.twitter.com/PhOL0F2UcY

— Barry Schwartz (@rustybrick) April 12, 2024

That made me think of this Gemini answer I got the day before, where Gary’s salt adjustment analogy could be related to “changes to the loss function or the parameters” of machine learning systems:

Backpropagation is a machine learning concept that (I feel) is relevant to today’s SEO lexicon. So far, this the best video I’ve seen explain it: https://t.co/K5jtKfSHty

I also asked Gemini why we might care generally, in light of Google’s core update. *LLM disclaimers apply. 😉 pic.twitter.com/JYkcxlBrlG

— Ethan Lazuk 🪬 (@EthanLazuk) April 12, 2024

Of course, I’m only positing that — 😉 — but I’d suggest studying backpropagation and the ML concepts around it.

As for when the core update rollout completes, that’s up to science. 🙂

Buckle up for a full week’s recap, and enjoy the vibes:

Thank you for supporting Hamsterdam and the cause of SEO & AI learning.

Missed last week? Don’t worry, I got you! Read Part 52 to catch up.

Other great sources of weekly SEO news:

• The SEO Weekly – Garret Sussman, iPullRank
• SEOFOMO – Aleyda Solis
• Weekly Video Recaps – Barry Schwartz, SER

---