Your company is considering deploying a chatbot, a “virtual customer assistant”, to initially handle customer service calls and, only when necessary, route callers to human operators for additional assistance. You’re discussing the options and concerns with your CMO (chief marketing officer) to determine how best to approach this project if your company decides to move forward with the initiative.

CMO:
I guess I'm a little old-fashioned, but I'm concerned this will actually be a turn-off for customers, that they'll get frustrated with the chatbot and we'll lose business.
You:
Today's AI systems present much more natural interactions than they used to. The system is learning as it goes and can even Selectsolve problemsmake mistakesself-correctauto-scaleItem 1 so its performance improves over time.

CMO:
That's amazing how smart computers are getting. How can customers know if they're talking to a computer or a real person?
You:
It's important to maintain transparency so customer expectations match reality. Still, we've not yet been able to develop a computer that could pass a(n) SelectIQWatsonTuringchatbotItem 1 test. There are always idiosyncrasies about how a computer communicates that you can detect, some kind of indication that the communication is not as natural as how a human talks.

CMO:
Does it really matter if you're talking to a computer, if it means you'll get the best response every time?
You:
Well, we're making a bit of a trade-off here. The kinds of problems we want computers to solve are too complex to write code that will always give the "right" or "best" answer. Sometimes we just need "good enough" in exchange for getting to the solution more quickly. Due to this complexity, AI systems rely heavily on SelectalgorithmsheuristicsimaginationsymbolsItem 1 .

You work for a security company that is developing a facial-recognition technology to help reduce shoplifting, organized retail crime, and return fraud. The idea is to incorporate facial recognition into existing security camera systems in stores, malls, airports, and sports venues that will identify people on wanted lists or previously convicted of criminal offenses (everything from shoplifting to violent crimes). The system then alerts security personnel of their presence along with advice to observe or to possibly alert authorities. This will serve a dual purpose of reducing crime-related losses and helping to keep customers and employees safe. The company is also considering incorporating the ability to identify missing persons through the software, such as abducted children or human trafficking victims.

The software will need to be able to learn as it goes, constantly improving its facial recognition performance. What kind of system can detect patterns in large sets of data that will enable this deep learning?

a. ANNb. DBaaSc. Vision systemd. Expert system

As the system identifies persons of concern, security personnel can help to further fine-tune the system's performance by providing feedback through the association of SelectdecisionsvaluespatternsconnectionsItem 1 to the results.

Your company is planning to incorporate AI into its manufacturing processes to leverage automation in the interest of greater productivity, efficiency, and worker safety. Employees are understandably concerned about how this might affect their work responsibilities or even their employment. You’re talking with the Operations manager about how best to support workers throughout this transition.

Operations manager:
Employees are worried. They don't want to lose their jobs, especially with the upcoming holidays. And I don't want to have to lay off anyone. But we've got to make this AI system work. How disruptive is all of this going to be?
You:
Research shows that human and AI collaboration is much more effective than using AI to replace humans. So ideally, we won't lose anyone. Instead, we can help employees SelectupskillspecializeplansupplementItem 1 to work with the AI instead of being replaced by it.

Operations manager:
Obviously, we can't send everyone back to school. What have you got in mind?
You:
There are a lot of options. For one, we can incorporate AI systems with easy-to-use Selectcustomizedpoint-and-clickcomplexcommand-lineItem 1 interfaces to democratize AI, making these systems easier for more people to learn how to use. We can also deploy lightweight robots where trained workers can decide how best to use these robots to complete strenuous, repetitive, or dangerous tasks.

Operations manager:
Wow, I didn't realize any of that was possible. So what kind of expert do we need on staff to make this happen and do it right?
You:
I think our best bet is a SelectML engineerSW developerAI research scientistrobotics scientistItem 1 . They can help us select user-friendly models, program them, and determine what kind of training our workers will need.

Machines must go through training to develop a foundation for further learning. Several types of machine learning are based on the kinds of data available for the process. These machine learning techniques can often be differentiated based on how much human users already know about the information they expect the machine to learn and output. For example, should the machine identify examples of a known category, or should the machine identify new categories of items by detecting previously undetected patterns of characteristics?

Identify which type of machine learning would be most helpful in each scenario.

Reinforced

Evelyn needs to train a system to determine the best timing for a network of traffic light signals.Supervised

Malia needs to train a system to detect indications of fraud using both known and unknown techniques based on the limited dataset of known fraud techniques.Semi-supervised

Leo needs to train a system to identify unexpected relationships between customer demographics and the products that they purchase.Unsupervised

Terrance needs to train a system to identify patients whose symptoms indicate various common diagnoses.

Robots in the industrial world are designed for speed, accuracy, and to keep workers safe in a fast-paced environment. Some are large models that work independently, and others are designed to work alongside humans. Robot designs are often categorized according to the types of movements the robot can make, the physical planes (such as horizontal and vertical) the robot can move within, and the amount of space (called a footprint) the robot takes up.

Identify which type of robot would be most useful in each scenario.

Cartesian

Terry needs a robot that will seal lids on cans as they pass by on a conveyor belt, which will require both lateral and rotary movement.Articulated

Gemini needs a robot that will pick up bunches of pencils from a flat conveyor belt and drop them into a nearby vertical funnel, which will require an adaptable grabbing motion followed by a large twisting motion.SCARA

Melody needs a robot that will automatically and precisely apply adhesive to metal cutouts. The robot should move within a cube-shaped space, similar to a claw machine at the arcade.

Answer rating: 100% (QA)

Evelyn will need a system that is able to switch different traffic lights on and off after a number of seconds As the machine is Supervised human oper... View the full answer