quit
Subscribe to daily Feed

Delivered by Google Feedburner...
Machine Learning

Machine Learning

posted 4 months 3 weeks ago
Product Analysis

Everyday we use thousands of different products, from telephones to bikes and drinks cans to washing machines. But have you ever thought about how they work or the way they are made?

Every product is designed in a particular way - product analysis enables us to understand the important materials, processing, economic and aestheticdecisions which are required before any product can be manufactured. An understanding of these decisions can help us in designing and making for ourselves.

Getting started

The first task in product analysis is to become familiar with the product! What does it do? How does it do it? What does it look like? All these questions, and more, need to be asked before a product can be analysed. As well as considering the obvious mechanical (and possibly electrical) requirements, it is also important to consider the ergonomics, how the design has been made user-friendly and any marketing issues - these all have an impact on the later design decisions.

Lets take the example of a bike:

What is the function of a bicycle?How does the function depend on the type of bike (e.g. racing, or about-town, or childs bike)?How is it made to be easily maintained?What should it cost?What should it look like (colours etc.)?How has it been made comfortable to ride?How do the mechanical bits work and interact?

If you do this exercise for various products, you will very quickly discover something interesting...

Systems and components

There are 2 main types of product - those that only have one component (e.g. a spatula) and those that have lots of components (e.g. a bike). Products with lots of components we call systems. For example:

ProductComponentsBikeFrame, wheels, pedals, forks, etc.DrillCase, chuck, drill bit, motor, etc.Multi-gymSeat, weights, frame, wire, handles, etc.

In product analysis, we start by considering the whole system. But, to understand why various materials and processes are used, we usually need to pull it apart and think about each component as well. We can now analyse the function in more detail and draft a design specification.

Some important design questions

To build a design specification, consider questions like the following:

What are the requirements on each part (electrical, mechanical, aesthetic, ergonomic, etc)?What is the function of each component, and how do they work?What is each part made of and why?How many of each part are going to be made?What manufacturing methods were used to make each part and why ?Are there alternative materials or designs in use and can you propose improvements?

These are only general questions, to act as a guide - you will need to think of the appropriate questions for the products and components you have to analyse. For a drinks container, a design specification would look something like:

provide a leak free environment for storing liquidcomply with food standards and protect the liquid from health hazardsfor fizzy drinks, withstand internal pressurisation and prevent escape of bubblesprovide an aesthetically pleasing view or image of the productif possible create a brand identitybe easy to openbe easy to store and transportbe cheap to produce for volumes of 10,000+

Once we have a specification, the next stage in the process is to understand how the materials are chosen.

Choosing the right materials

Given the specification of the requirements on each part, we can identify the material properties which will be important - for example:

RequirementMaterial Propertymust conduct electricityelectrical conductivitymust support loads without breakingstrengthcannot be too expensivecost per kg

One way of selecting the best materials would be to look up values for the important properties in tables. But this is time-consuming, and a designer may miss materials which they simply forgot to consider. A better way is to plot 2 material properties on a graph, so that no materials are overlooked - this kind of graph is called a materials selection chart (these are covered in another part of the tutorial).

Once the materials have been chosen, the next step is normally to think about the processing options.

Choosing the right process

It is all very well to choose the perfect material, but somehow we have to make something out of it as well! An important part of understanding a product is to consider how it was made - in other words what manufacturing processes were used and why. There are 2 important stages to selecting a suitable process:

Technical performance: can we make this product with the material and can we make it well?Economics: if we can make it, can we make it cheaply enough?

Process selection can be quite an involved problem - we deal with one way of approaching it in another part of the tutorial.

So, now we know why the product is designed a particular way, why particular materials are used and why the particular manufacturing processes have been chosen. Is there anything else to know?

Final remarks

Product analysis can seem to follow a fixed pattern:Think about the design from an ergonomic and functional viewpoint.Decide on the materials to fulfil the performance requirements.Choose a suitable process that is also economic.

Whilst this approach will often work, design is really holistic - everything matters at once - so be careful to always think of the bigger picture. For example:

Is the product performance driven or cost driven? This makes a big difference when we choose materials. In a performance product, like a tennis racquet, cost is one of the last factors that needs to be considered. In a non-performance product, like a drinks bottle, cost is of primary importance - most materials will provide sufficient performance (e.g. although polymers arent strong, they are strong enough).Although we usually choose the material first, sometimes it is the shape (and hence process) which is more limiting. With window frames, for example, we need long thin shaped sections - only extrusion will do and so only soft metals or polymers can be used (or wood as it grows like that!
Like | 6 View | 602

Machine Learning

posted 7 months 1 day ago

 

Allo, Google’s AI-enhanced answer to ‘smart messaging’ on mobile, is here. Designed to keep users from straying outside the app to search for things on the internet, Allo is the first Google product to feature its AI “assistant.”
 
Suggestions from the assistant are meant to be conversational, and can be plugged into existing conversations or had between you and Google alone. Google is expected the start rolling out the assistant to other products this fall, starting with Google Home, as part of the company’s ongoing push toward AI. Allo comes on the heels of Duo, Google’s video calling app, which has been downloaded 10 million times since it was released last month. Taken together, the two represent Google’s attempt to grab some of the direct messaging market while integrating machine learning across its suite of consumer-oriented products. “We don’t see messaging as a solved problem,” said Nick Fox, Google’s vice president of communications products. Fox said Allo is about “getting things done right in your chat. We think the enabler here is AI.”
 
Fox stressed that the goal of Allo is to keep the automated suggestions simple and subtle, so as not to replace other apps or search generally, but rather to supplement them. Like Duo, Allo uses your cell phone number, so there’s no need to create a separate user account. Allo does associate with your existing Google account, however, giving it access to a host of personal information, such as images you’ve saved with Google’s cloud photo storage. The more you use the assistant, the more it learns about you. Once you tell Allo your favorite sports team, for example, you can recall news about the team without using its name. “Google has been a one on one experience for 18 years,” Fox said, adding that with Allo, “now it’s like multiplayer.”
 
 
 
When you turn Allo on, it asks for your location. This gives it the ability to search for things you might be likely to ask it for, such as the weather or nearby restaurants. Allo retains the context of a conversation when you query it , mimicking an actual conversation you’d have with a friend. Let’s say you invite someone to dinner via Allo, and your friend asks the assistant to find nearby restaurants. Both users would see the same results, and if one person wanted to see just the restaurants that are open or those with the highest rating, for example, the assistant would filter down results accordingly, all within the app itself.
 
The assistant gives two types of results. The first is what you’ve come to expect from any search engine, and includes basic information on the subject you’ve asked about. Beneath that is a row of suggested information based on what you asked, when you asked and what you’ve asked in the past. Throughout the app, results are doled out in “bite size snacks,” Fox said. Think of results less like a comprehensive Wikipedia page, and more direct responses to the question you’ve just asked. In addition to quickly surfacing web results, Allo lets you respond to messages with pre-determined phrases that are common replies to questions or prompts. So if your friend sends you a selfie, for example, an automated reply might be “What a great smile.”
 
The risk of any smart messaging system, Fox said, is that it gets in the way of the conversation. Google’s goal is to have its assistant be as unobtrusive as possible, while still adding an element of ease and simplicity that people have come to expect. “The user is in control,” he said. “That’s a theme throughout.”
 
Among the other features of Allo: Photos sent through the app fill up more than half your screen, and appear edge to edge width on your phone rather than in their own message bubble that needs to be opened. Allo’s “whisper” or “shout” functions allow you to control the size of text or emojis. In addition to all the standard emojis you’re familiar with, you can download a slew of “sticker packs,” featuring cartoon-like images created by artists Google has contracted with. Allo’s “incognito” mode features end-to-end encryption, and allows you or the person you are messaging with to set messages to expire. When set to incognito, Allo defaults to discreet notifications that don’t include the sender’s identity or content of the message until the app has been opened on your phone.
 
For now, the assistant won’t share personal information–such as photos you’ve stored in the cloud–with contacts you are messaging, though this will likely happen later as Google tweaks with limits on the type of information that’s shared. The assistant, which Google says it will begin rolling out in other products this fall, is in “preview” edition, and will prompt you for feedback on individual chats. If the person you are messaging doesn’t have the app installed, the message is still delivered via standard SMS text. The sender will get a notification saying the recipient doesn’t have Allo, and the recipient will see a link to download the app. Allo is available for both Android and iOS.
Like | 3 View | 604