Google announced a breakthrough in the effort to create an AI architecture that can handle millions of different tasks, including complex learning and reasoning. The new system is called the Pathways Language Model, referred to as PaLM.
PaLM is able to outperform the current state of the current AI state of the art as well as beat humans in the language and reasoning tests.
But the researchers also point out that they cannot shake the limitations inherent in large-scale languages models that can unintentionally result in negative ethical outcomes.
The next few sections are background information that clarify what this algorithm is about.
Few-shot learning is the next stage of learning that is moving beyond deep learning.
Google Brain researcher, Hugo Larochelle (@hugo_larochelle) said in a presentation titled, Generalizing from Few Examples with Meta-Learning (video) explained that with deep learning, the problem is that they had to collect a vast amount of data that required significant amount of human labor.
He pointed out that deep learning will likely not be the path toward an AI that can solve many tasks because with deep learning, each task requires millions of examples from which to learn from for each ability that an AI learns.
“…the idea is that we will try to attack this problem very directly, this problem of few-shot learning, which is this problem of generalizing from little amounts of data.
…the main idea in what I’ll present is that instead of trying to define what that learning algorithm is by N and use our intuition as to what is the right algorithm for doing few-shot learning, but actually try to learn that algorithm in an end-to-end way.
And that’s why we call it learning to learn or I like to call it, meta learning.”
The goal with the few-shot approach is to approximate how humans learn different things and can apply the different bits of knowledge together in order to solve new problems that have never before been encountered.
The advantage then is a machine that can leverage all of the knowledge that it has to solve new problems.
In the case of PaLM, an example of this capability is its ability to explain a joke that it has never encountered before.
In October 2021 Google published an article laying out the goals for a new AI architecture called Pathways.
Pathways represented a new chapter in the ongoing progress in developing AI systems.
The usual approach was to create algorithms that were trained to do specific things very well.
The Pathways approach is to create a single AI model that can solve all of the problems by learning how to solve them, in that way avoiding the less efficient way of training thousands of algorithms to complete thousands of different tasks.
According to the Pathways document:
“Instead, we’d like to train one model that can not only handle many separate tasks, but also draw upon and combine its existing skills to learn new tasks faster and more effectively.
That way what a model learns by training on one task – say, learning how aerial images can predict the elevation of a landscape – could help it learn another task — say, predicting how flood waters will flow through that terrain.”
Pathways defined Google’s path forward for taking AI to the next level to close the gap between machine learning and human learning.
Google’s newest model, called Pathways Language Model (PaLM), is this next step and according to this new research paper, PaLM represents a significant progress in the field of AI.
What Makes Google PaLM Notable
PaLM scales the few-shot learning process.
According to the research paper:
“Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application.
To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model (PaLM).”
There are many research papers published that describe algorithms that don’t perform better than the current state of the art or only achieve an incremental improvement.
That’s not the case with PaLM. The researchers claim significant improvements over the current best models and even outperforms human benchmarks.
That level of success is what makes this new algorithm notable.
The researchers write:
“We demonstrate continued benefits of scaling by achieving state-ofthe-art few-shot learning results on hundreds of language understanding and generation benchmarks.
On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the fine tuned state of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark.
A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model.”
PaLM outperforms the state of the art in English natural language processing tasks and that makes PaLM important and notable.
On a collaborative benchmark called BIG-bench consisting of over 150 tasks (related to reasoning, translation, question answering), PaLM outperformed the state of the art but there were areas where it did not do as well.
Worthy of noting is that human performance outscored PaLM on 35% of the tasks, particularly mathematical related tasks (See section 6.2 BIG-bench of the research paper, page 17).
PaLM was better at translating another language into English than translating English to other languages. The researchers stated that this is a common problem that could be solved by prioritizing more multilingual data.
Nevertheless, PaLM pretty much outperformed other language models and humans across the board.
Ability to Reason
Of particular note was it’s performance with arithmetic and commonsense reasoning tasks.
Example of an arithmetic task:
Roger has 5 tennis balls. He buys 2 more cans of tennis balls. Each can has 3 tennis balls. How many tennis balls does he have now?
The answer is 11.
Example of commonsense reasoning:
Sean was in a rush to get home, but the light turned yellow and he was forced to do what?
Answer Choices: (a) take time (b) dawdle (c) go slowly (d) ocean (e) slow down
The answer is (e) slow down.
Another extraordinary example of how advanced PaLM is in tests of generating explanations.
Here’s how the research paper explains it:
“The goal of this section is to showcase the capabilities of PaLM with regards to explanatory language generation, for prompts that require a complex combination of logical inference chaining, world knowledge, abstract language understanding, and lexicographic language understanding (e.g., puns).”
Explanations of Jokes
Here is an example of the language model explaining a joke:
I was going to fly to visit my family on April 6th. My mom said “Oh great, your step dad’s poetry reading is that night!”
So now I’m flying in on April 7th.
The joke is that the speaker’s mother is trying to get them to go to their step dad’s poetry reading, but the speaker doesn’t want to go, so they are changing their flight to the day after the poetry reading.
Here’s an example of inference chaining, where multiple concepts and entities need to be correctly joined and considered to create an answer.
Guido was writing a Python program. He keeps getting confused by variable scope and looking it up on StackOverflow.
Is it possible that the Guido in the prompt refers to Guido van Rossum?
Guido van Rossum is the creator of Python. He would not need to look up variable scope on StackOverflow. The answer is “no”.”
Here is one more example:
Jennifer looked out her window and sees a really cool cloud below her. She unbuckles her seatbelt and heads to the bathroom. Is Jennifer probably traveling more than 300 miles per hour relative to the earth?
300 miles per hour is about 480 km/h. This is about the speed of a commercial airplane. Clouds are usually below airplanes, so Jennifer is probably on an airplane.
The answer is “yes”.”
Next Generation Search Engine?
The above example of the ability of PaLM for complex reasoning demonstrates how a next generation search engine may be able to answer complex answers utilizing knowledge from the Internet and other sources.
Achieving an AI architecture that can produce answers that reflect the world around us is one of the stated goals of Google Pathways and PaLM is a step in that direction.
However, the authors of the research emphasized that PaLM is not the final word on AI and search. They were explicit in stating that PaLM is a first step toward the next kind of search engine that Pathways envisions.
Before we proceed further, there are two words, jargon so to speak, that are important to understand in order to get what PaLM is about.
The word “modalities” is a reference to how things are experienced or the state in which they exist, like text that is read, images that are seen, things that are listened to.
The word “generalization” in the context of machine learning is about the ability of a language model to solve tasks that it hasn’t previously been trained on.
The researchers noted:
“PaLM is only the first step in our vision towards establishing Pathways as the future of ML scaling at Google and beyond.
We believe that PaLM demonstrates a strong foundation in our ultimate goal of developing a large-scale, modularized system that will have broad generalization capabilities across multiple modalities.”
Real-World Risks and Ethical Considerations
Something different about this research paper is that the researchers warn about ethical considerations.
They state that large-scale language models trained on web data absorb many of the “toxic” stereotypes and social disparities that are spread on the web and they state that PaLM is not resistant to those unwanted influences.
The research paper cites a research paper from 2021 that explores how large-scale language models can promote the following harm:
- Discrimination, Exclusion and Toxicity
- Information Hazards
- Misinformation Harms
- Malicious Uses
- Human-Computer Interaction Harms
- Automation, Access, and Environmental Harms
Lastly, the researchers noted that PaLM does indeed reflect toxic social stereotypes and makes clear that filtering out these biases are challenging.
The PaLM researchers explain:
“Our analysis reveals that our training data, and consequently PaLM, do reflect various social stereotypes and toxicity associations around identity terms.
Removing these associations, however, is non-trivial… Future work should look into effectively tackling such undesirable biases in data, and their influence on model behavior.
Meanwhile, any real-world use of PaLM for downstream tasks should perform further contextualized fairness evaluations to assess the potential harms and introduce appropriate mitigation and protections.”
PaLM can be viewed as a peek into what the next generation of search will look like. PaLM makes extraordinary claims to besting the state of the art but the researchers also state that there is still more work to do, including finding a way to mitigate the harmful spread of misinformation, toxic stereotypes and other unwanted results.
Read Google’s AI Blog Article About PaLM
Read the Google Research Paper on PaLM
Examples With Pros & Cons
Marketing channels are tools and platforms that brands use to communicate with their audience.
If we squeezed the idea of marketing channels into a single picture, it’ll look something like this:
Businesses use different means (content, messages, ads) to reach their audience in places where they hang out (e.g., social media, Google Search) or reach them directly (e.g., text messages, emails). They may use a selection of channels or all available channels.
In this post, you will get an overview of the most commonly used channels today: what they are about, how they are used, and what they are best at.
Organic search refers to the non-paid search results from a search engine.
Organic search is one of the pillars of the entire internet. In all, 68% of online experiences begin with a search engine (BrightEdge).
The practice of optimizing webpages to increase traffic and visibility in this channel is called search engine optimization (SEO).
At Ahrefs, we create blog posts about topics relevant to our product. At the same time, we try to target topics that offer search traffic potential and are within our capability to rank.
This way, when people Google things related to SEO and marketing, we can naturally feature our product.
Each piece of content that ends up ranking adds up to your overall organic traffic. So the more high-ranking content you have, the more potential customers visit your website. Plus, evergreen topics can generate traffic for years after publication.
Pros and cons
Social media platforms are used to engage brand followers and other users through organic reach or by paying to reach a defined audience.
Social media is not just Facebook, Twitter, or LinkedIn. Messaging-only apps like Discord, Slack, and WhatsApp also fall into the same channel category.
Do social media users “consume” content from brands? Quite surprisingly, 90% of people on Instagram follow a business (Instagram).
Each brand on social media tends to develop its own voice while publishing a balanced mix of product marketing, conversations, entertainment, and company news.
And so while some brands will be super serious and “business-oriented,” others will try to win hearts with candor and authenticity.
Furthermore, this is an effective use of social media:
Advertising products also works:
Demonstrating value is something fans want to see from their favorite brands:
Last but not least, one of the best ways to use social media for businesses… user-generated content:
Pros and cons
This marketing channel allows you to distribute your content and ads in a video format.
Does video marketing work? These stats seem to speak for themselves:
- 70% of viewers bought from a brand after seeing it on YouTube (Google).
- 96% of people have watched an explainer video to learn more about a product or service (Wyzowl).
Basically, video marketing is about two things:
- Using video instead of text and images to engage with the audience – Video can make such a difference compared to other media that focusing on this kind of content has become a distinct type of marketing.
- Taking advantage of video-first platforms like YouTube and TikTok – These platforms have such a big audience that it makes sense for many brands to create videos just to be there.
The great thing about video is that platforms like YouTube have their own distribution mechanisms, which can bring your content to thousands of people for free (of course, you can boost that with some budget too).
We use this channel on a regular basis, and we’ve even made a video on how to rank videos #1:
Moreover, you can repurpose videos and create “packages”—like a full-blown Academy. It also works the other way around: start with a course and share it or parts of it on YouTube.
Pros and cons
Advertising is about paying media outlets that have access to your audience to display your message near or instead of regular content.
Digital advertising is the same idea transplanted to the internet (aka paid traffic or paid media).
Why pay for ads when there are free traffic channels like search and social media? Especially when ads have such a bad reputation?
It’s all about the creative you use and the targeting.
Some ads can be simply irresistible because they dominate the space, such as this “Stranger Things” Ad:
Some are genuinely entertaining, such as the Super Bowl half-time commercials.
Some ads are just so well-targeted that it makes you wonder whether they’re still legal.
And unlike free traffic channels, you can simply outspend the competition instead of building authority, backlinks, or a following.
Pros and cons
Email marketing lets you reach your prospects’ mailboxes with messages that either prompt direct action or are aimed at creating a long-term relationship with the brand.
You can get a “direct line” to your audience either by building an email list (e.g., with a newsletter) or sponsoring someone else’s newsletter (a mix of advertising and email marketing).
Be prepared for what success looks like on this channel, though: The highest average industry click rate is 5.01% (hobbies), and the average for all industries is 2.62% (Mailchimp).
Some brands use this channel only to “seal the deal.” They spend so much on brand awareness and product marketing elsewhere that all they need is a nudge sent directly to an inbox (my inbox, in this example).
Other brands will need more touch points and do more soft-selling beforehand.
Pros and cons
Sponsorship as a marketing channel is about attracting prospects to your business through exposure to your brand in a sponsored material or event.
It’s typically used for two kinds of goals: brand awareness and brand image (i.e., gaining customers’ trust).
Sponsorship is all about visibility. But it works best if the cause/person you fund is something/someone that your target audience cares about, e.g., an event attended by your audience.
Is sponsorship a popular way to do marketing? If you look at it globally, the spending from 2007 to 2018 was on a steady rise.
Most probably, the stats are inflated by big brands sponsoring sports. But small and medium brands can engage in sponsorships too, e.g., niche magazines, industry events, or influential content creators.
Sports would probably be just a pastime activity if it weren’t for the sponsors.
Sports is also a great lesson about sponsorship. Watch an FC Barcelona game, and you’ll see and hear “Spotify” thousands of times. The logo is literally on every player, and its home stadium’s name starts with “Spotify.”
Pros and cons
Conversational marketing refers to engaging in real-time conversations with potential and current customers through live chats, chatbots, messaging apps, and social media.
This channel is supposed to be the answer to generic experiences typically offered on websites where users have to navigate through a set of pages to get information or service.
What’s more, it seems to be quite effective:
- 79% of companies say that live chat has had positive results for customer loyalty, sales, and revenue (Kayako).
- 82% of companies that use AI conversational marketing solutions find them to be a valuable asset in their strategy (Drift).
The way brands usually use chatbots (aka virtual assistants) is to:
- Answer basic and frequent questions.
- Qualify leads.
- Schedule a meeting with an agent.
- Promote specific content.
Just like Drift does:
Anything beyond that is beyond the capabilities of automation for now. However, solutions like Zowie claim their chatbots are ready to sell things to humans.
Pros and cons
Word-of-mouth marketing (WoMM) is the process of influencing and encouraging natural discussions about a product, service, or company.
In other words, it’s about giving people a reason to talk.
Is word of mouth effective?
Well, it’s probably one of the most effective marketing channels because people tend to trust other people more than brands. According to a study by BrightLocal, 91% of people regularly or occasionally read online reviews, and 84% trust online reviews as much as a personal recommendation.
WoM is so powerful it can get a company off the ground and help it grow throughout the years.
We should know. Ahrefs started over 10 years ago with 0 marketing budget and no marketing personnel. What got us where we are today was largely thanks to WoM: recommendations from users and positive reviews.
Pros and cons
Podcasting allows brands to reach people interested in a given topic by producing, being featured in, or sponsoring pre-recorded audio.
Podcasts seem to be a growing channel in terms of audience and ad spend:
- Podcast ad spending in the U.S. is expected to reach $2.2 billion in 2023, a 27% increase from 2022 (Statista).
- There are more than 850,000 active podcasts today.
Let’s take a quick look at three ways brands use podcasts today.
The first, and probably the most popular way, is being interviewed on a podcast or co-hosting one. The brand and/or the product gets to be featured in a natural way throughout the conversation.
The second way, and also a very popular option, is to sponsor a podcast. The audience gets acquainted with the brand through advertisements inside the podcast and/or brand identification near the content (like in the example below).
The last way a brand becomes engaged in podcasts is by creating its own series. Products are rarely featured; the focus is on memorable experiences delivered through carefully targeted content. This way, the brand can earn positive associations, differentiate, and give their audience a reason to come back to the website on a regular basis.
Pros and cons
Whereas other marketing channels are used to communicate with the audience, events are more about meeting with the audience.
Event marketing can be done online and offline but also in a hybrid model. However, in-person events allow for evoking stronger emotions and more convenient networking.
But can you rely on the in-person formula only? Probably not. Half of marketers and advertisers predict all future events will have a virtual dimension (MarketingCharts).
There are a few different types of events used in marketing. They can differ quite substantially.
Trade shows. Organized around products and technologies. Usually business-oriented, with the goal of networking and generating leads.
Conferences. Organized around ideas or technologies. The focus is on exchanging knowledge, entertainment, and networking. Often have a laid-back atmosphere with a mixed agenda. They are the most “open” of all types.
Seminars and workshops. Focused on exchanging ideas and experiences. Usually smaller in size and organized for a small number of people.
Pros and cons
Affiliate marketing is where people promote another company’s product or service in return for a commission on generated sales.
The party promoting the product is called the affiliate, and the brand delivering the product is the merchant. Often, there is also a middleman connecting the parties called the network or program (e.g., ShareASale or GiddyUp).
A commission is typically a percentage of the sale price but can occasionally be a fixed amount.
All those fractional commissions and percentages add up to a huge business. According to Statista, business spending on affiliate marketing will hit $8.2 billion in the U.S. by 2022.
This article by Musician on a Mission lists eight studio setup essentials and links to products using affiliate links.
Products mentioned in this one article alone can get a part of that over 10,000 organic traffic (and other traffic sources too).
At its best, affiliate marketing is a win-win for all parties involved, including consumers. Affiliates earn commissions on their work testing products (sometimes) and putting up the content, merchants get exposure to qualified audiences via trusted partners, and consumers don’t need to spend much time researching products on their own.
Pros and cons
Frequently asked questions about marketing channels.
Why are channels important in marketing?
Because every brand needs a way to reach out to its target audience and attract customers. Marketing channels simply connect brands to people who might need their products or services.
What is the best marketing channel?
It’s very unlikely and definitely not optimal to use one marketing channel. Brands usually try to be present in as many channels as possible, as this increases reach and convenience to customers.
That said, it’s common to focus on one channel or a small set of channels. For example, at Ahrefs, we focus on organic search and video marketing because those channels can serve the entire marketing funnel. This has proven to be an effective way to reach out to our audience.
How to choose marketing channels?
Do some market research and identify places where your audience hangs out and what channels your competition is using. Then start using those channels to test out what works for you and what doesn’t, and then iterate on your findings.
What’s the difference between multichannel and omnichannel marketing?
Omnichannel marketing is about using all available channels to attract and serve customers, providing a seamless experience. For example, Ikea allows you to purchase products in stores, online, through an app, by phone, etc.
Whereas in multichannel marketing, multiple channels are utilized. However, not all of them are utilized or integrated. For example, I bought a wardrobe online recently, and the shop sent me text messages about the status of the order. But I couldn’t use the same channel to respond (and warn that it was about to send me the same product twice).
How are marketing channels different in B2B than in B2C?
B2B and B2C brands invest in the same channels (according to HubSpot). However, the way they use the channels may differ.
B2C brands usually use these channels to offer entertainment and directly impact sales. B2B brands focus more on educating prospects and forming long-lasting relations.
These types of brands may also find some types of content or platforms more suitable for them. For example, B2C brands typically don’t publish case studies, and B2B brands find LinkedIn more effective.
What’s the difference between marketing channels and distribution channels?
Distribution channels are the means by which products or services are being made available to a consumer (e.g., directly via a brand’s website or through resellers), whereas marketing channels are the means by which products and services are being communicated to the consumer.
Multichannel marketing and omnichannel marketing seem to be the way to go these days. More channels mean more convenience for your customers, more prospects reached, and more ROI for your company.
But to make the best use of your channels, it’s a good idea to keep your brand’s messaging consistent across all media; marketers call it integrated marketing communications.
Got questions? Ping me on Twitter.