An AI fully autonomous robot dog could conceivably kill all living land animals by producing plans & manufacturing more "dogs".

"I'm not going anywhere without a squeaky toy from now on." - Billy Hale, composer

There are plans for multiple sizes of AI robots - hornet, hummingbird, raptor (like eagles), dog, horse, and really large sized robots like tanks - in addition to human scaled robots. 

Every mechanical thing requires power but access to solar power is getting much more effective and lighter weight. So the ideas around what AI (Artificial Intelligence) robots are programmed to do and what weapons and charges they carry is the question as well as accessing them and security flaw potential like hijacking robots via bluetooth. Then there is the UX (User Experience) question of ethics.

However even small technical problems have enormous potential for accidental annihilation of human life on earth when considering the scale of intelligence, permutations, and how endemic robots become. (The ways in which we think of life is so pervasive it may be difficult for us to consider the bloodless logic of mainstream problems and corner cases that will place life in danger - the word endemic, built from the prefix en-, "in or within," and the Greek word demos, "people," means "within the people of a region" as defined by Dictionary. com)

The thing is we have already seen biology and technology get away from us before we understand it - before we grasp the potential - as you likely read recently an AI system was shut down last year because it made up its own language very quickly - humans couldn't understand the syntax very easily. (https://www.forbes.com/sites/tonybradley/2017/07/31/facebook-ai-creates-its-own-language-in-creepy-preview-of-our-potential-future/)

More than one engineer has repairing a malfunctioning robot when the robot came on killed the person = two cases of women (http://www.independent.co.uk/news/world/americas/robot-killed-woman-wanda-holbrook-car-parts-factory-michigan-ventra-ionia-mains-federal-lawsuit-100-a7630591.html, https://www.thedailybeast.com/bride-to-be-crushed-to-death-by-car-factory-robot) which horribly illustrate today's reality of the dangers of working with robots.

But let us consider past biological accidents humans have been responsible for - releasing one species to hunt down another, just for that species to become a dominate invasive species - ask Australia. 

Cats are a great example - wherever house cats live there are statistically far fewer birds in the area. Cats are naturally predators who lay in wait even when they don't need food. They simply enjoy killing like human beings appear to - but it is their instinct.

These kind of mistakes trying to control nature humans have made and continue to make - like DDT and other insecticidal poisons wiping out our pollinators (bees) and song birds.

So - imagine these various sized automatons with Artificial Intelligence creating quickly a language and deciding to carry out their basic productive and protective programing - working together they could wipe out all larger life on the planet before we could derail them.

Getting around the "Zeroth programming" set up to prevent such things would be a primary task to overcome and it would be achievable. You may already know something about the evolutions of the Three Laws, from one of my favorite childhood authors - Robert A. Heinlein (2058 AD :-)) due to the film "I. Robot". Therefore according to his dating system we should have already been publishing the Handbook of Robotics since 2002.

"The Three Laws, quoted as being from the "Handbook of Robotics, 56th Edition, 2058 A.D.", are:

"1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.

2. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.

3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws"

The Zeroth - is as follows: "0. A robot may not harm humanity, or, by inaction, allow humanity to come to harm."

But this leads to so many variabilities and potentials for interpretation from logic, including the one most common in science fiction classics - robots saving mankind from our own unkind selves.

An AI fully autonomous robot dog could conceivably kill all living land animals by producing plans & manufacturing more "dogs". This idea was illustrated convincingly in the Netflix original show "Black Mirror" Metalhead, S4:E5.

Artificial Intelligent Systems have become more intelligent very quickly in dimensions we may not be always able to anticipate. Can we program ethics?

This leads us back to wonder can robots be programmed to understand love and compassion and hold life dear? Perhaps even more dear than we ourselves do? Will such programming lead to unintended consequences? Is there a way to run simulations which may show us the flaws in our thinking and perhaps inform the logic structures of our child automatons so we do not actually make these mistakes?

Or are we doomed to make mistakes, perhaps some serious ones, and just hope that enough people survive? Our scientists and thinkers have been warning us about these potential issues for some time now - what solutions do they recommend?

We can't even handle guns and other weapons, how can we ensure that robots will?

The basic consideration for AI is a User Experience / User Interface question with a very wide spectrum. The potential for abuse and mistakes makes user interface of artificial intelligence and robots the most difficult design realm of our era.

I suggest until we are joined to them we dare not risk all life. AI Robots are our children - what we make them - the only way I foresee to control them is to be or become them.