T wo largely unquestioned assumptions long ago took root within the space community. As we prepare to voyage back to Earth’s Moon and on to Mars, it is time to question them both .

The first assumption is that exploring the Moon, Mars, or any part of the solar system, can be accomplished in a generation or two and with limited loss of life. The second is that we can use robots to successfully understand another world. Both assumptions are almost certainly wrong, yet many important elements of our civil space program are based on one or both of them being correct.

To paraphrase Douglas Adams, even within the space community most people don’t have a clue how “mind-boggingly big space really is.” Most of the major worlds in the solar system have surface areas at least as large as terrestrial continents — a few are much larger — and every one of them is unremittingly hostile to human life. Learning to travel confidently through former President John F. Kennedy’s “this new ocean” will be difficult, expensive, time-consuming and dangerous.

Mr. Kennedy’s rhetoric was more accurate than he probably knew. The only remotely comparable task humanity has faced was learning to travel across our world’s oceans. We take trans-oceanic travel for granted, but getting from Neolithic boats to modern freighters cost humanity well over 10,000 years of hard work and uncounted lives. Even today, hundreds of people die in shipping accidents every year. We and our woefully inadequate chemical rockets are like Stone Age tribesfolk preparing to cast off in canoes, reaching for barely visible islands over a freezing, storm-tossed, North Atlantic.

The salient fact is, while it was much more difficult than most people care to remember, we did learn to ply the oceans, even arctic ones.

Space is harder still, but many of the problems are similar: the environment is alien and deadly and most supplies must be carried along. Like our Neolithic friends, we can see our destinations in the distance.

With Apollo, we visited the closest island and a series of progressively more sophisticated space stations has demonstrated long-term survival in the shoals close to home.

The task may be far harder than we imagine, but it is not impossible.

The second assumption is that we can conduct detailed exploration with robots, without personal risk or people on site.

Recent events should engender some humility in our toolmakers. In spite of all the money spent on space robotics and some extraordinary successes like the Mars Exploration Rovers, we have failed to reliably automate even relatively simple tasks.

Docking two spacecraft together would seem an ideal job for automation, but recent experiments such as the Demonstration of Autonomous Rendezvous Technology have not gone well. The Russian masters of this skill keep well-trained cosmonauts in reserve at the space station whenever possible, and they have to take over with depressing regularity.

If we cannot reliably automate docking in Earth orbit, what makes us think we can do so at Mars as part of an expensive effort to return a few small samples?

We recently learned it would cost at least as much to automate the repair of the Hubble Space Telescope, with a lower chance of success, than to do so with astronauts — even when the latter used the expensive space shuttle. Closer to home, an Air Force audit recently discovered, contrary to expectations, that it costs more to run automated spy planes than it does the human-piloted variety.

We cannot cheaply or reliably automate the use and repair of well-understood nearby machines, with known interfaces between parts and tools. Why do we think a robot could, say, find a fossil on the rugged, random and largely unknown landscape of Mars? Finding a fossil on Earth requires scouting wide areas for likely rocks. You must hold and handle many thousands of oddly shaped samples of different size and weight, and with differing cohesions and textures; and observe all of them from any angle and at any scale.

It also involves being able to cleanly cut samples of many sizes along any axis; examining each cut at a wide range of scales and wavelengths; and doing sophisticated on-the-fly pattern recognition to recognize any fossil. No foreseeable robot, at any cost, will be able to simultaneously handle any combination of these tasks. A single geologist with a limited set of tools can quickly do them all.

Finding the second fossil will be no easier, nor will the third or fourth; then we need to study their distribution, and their positions within layers of accurately dated rock up to kilometers deep, to determine the fossil’s age and history. It is barely conceivable we could automate the detection of life on Mars. Understanding any life, or ruling out life’s existence, requires scientists on site.

Many of the same issues apply when attempting to understand the fine-scale layering and dating of lunar volcanic flows or Martian sediments.

Our rovers’ accomplishments on Mars are remarkable and exciting, but let’s not lose perspective and inflate their achievements. The rover project has spent more than two Earth years and well over $1 billion traveling less distance than human geologists could walk in an afternoon. The rovers helped us discover that, at some undetermined date in the past, there was standing water on Mars. It is no disrespect to one of the great accomplishments of our age to point out that this is basic reconnaissance with precious little science.

On the other hand, we have consistently underestimated the science achieved by human expeditions. The only absolute dating of any surfaces in the solar system, other than those on Earth, was obtained on the Moon by Apollo astronauts.

From Mercury to Neptune’s moons, the dates derived from crater counts are little more than educated guesses relative to the Apollo record.

Historian Asif A. Siddiqi, in his seminal history of the Soviet lunar program “Challenge to Apollo” argues that it is not clear the random 105 gram sample collected by the Soviet Luna-16 robot was scientifically more cost effective than the Apollo samples obtained by the two crews flown by the same date. Apollos 11 and 12 returned an intelligently collected, well-documented 60 kilograms obtained across wide areas.

“Luna-16 was certainly a remarkable technological accomplishment,” Siddiqi wrote, “but it was probably not, as Soviet officials of the day touted, a ‘cheaper and better’ alternative to Apollo.”

Dramatic increases in exploration funding are not likely in the foreseeable future. If we are going to make progress toward truly understanding the Moon and Mars, we must send scientists while staying close to existing budgets. Whatever the dangers, we must proceed with our existing tools and technologies.

Dangerous it will be. Detailed exploration, let alone settlement, of nearby worlds will be the single most difficult task humanity has ever tackled. Most likely, it will take many hundreds, or even thousands, of years. Our first attempts to establish a base on Earth’s Moon or Mars may well fail. As on the oceans, many people will die: we cannot insist on levels of safety that make the exercise technically impractical or unaffordable.

Like those Neolithic sailors setting out across ice-flecked waves in dugouts, if we wait for the perfect tools we will never go. Although they can barely do the job, shuttle-derived chemical rockets are what we have, so we must use them. Better ships will come as we create reasons for them to be developed, such as early scientific bases on other worlds that require regular deliveries of supplies.

NASA’s relatively affordable and practical plan to return astronauts to Earth’s Moon using available technology is the best, and probably only, way forward.

Is it worthwhile to send people across this new ocean to the planets? Was it worth the resources and lives to learn to send people across our own world’s oceans to distant islands and unknown continents? Neither question has a simple or entirely logical answer, but in both cases any answers are the same.

Donald F. Robertson is a freelance space industry journalist based in San Francisco.